141 61 40MB
English Pages 208 [209] Year 1998
Unity in Diversity? Current Trends in Translation Studies
Edited by
Lynne Bowker, Michael Cronin, Dorothy Kenny and Jennifer Pearson
First published 1998 by St Jerome Publishing Published 2014 by Routledge 2 Park Square, M ilton Park, Abingdon, Oxon 0X 14 4RN 711 Third Avenue, New York, NY 10017, USA Routledge is an imprint o f the Taylor & Francis Group, an informa business Copyright © 1998, Taylor & Francis All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. ISBN 13: 978-1-900650-15-1 (pbk)
British Library Cataloguing in Publication Data A catalogue record of this book is available from the British Library Cover Illustration: St. Jerome, by Nicolas Frankes Courtsey of The National Gallery of Ireland Cover design by Steve Fieldhouse, Oldham, UK
Contents Introduction Section One: The Nature of Translation Luise von Flotow Dis-unity and Diversity Feminist Approaches to Translation Studies Susan Ingram Translation, Autobiography, Bilingualism Section Two: Translation in National Context Ian Brown and Ceri Sherlock Antigone A Scots/Welsh Experience o f Mythical & Theatrical Translation Aniko Sohar ‘Genuine ’ and ‘Fictitious ’ Translations o f Science Fiction and Fantasy in Hungary Paul St.-Pierre Theory and Practice: Translation in India Judy Wakabayashi Marginal Forms o f Translation in Japan Variations from the Norm Eithne O’Connell Choices and Constraints in Screen Translation Section Three: Descriptive Translation Studies Irena Kovacic Six Subtitlers - Six Subtitling Texts Christina Schaffner Parallel Texts in Translation Carol Peters and Eugenio Picchi Bilingual Reference Corpora fo r Translators and Translation Studies
Sara Laviosa The English Comparable Corpus A Resource and a Methodology
101
Section Four: Computer-Aided Translation Sharon O’Brien Practical Experience o f Computer-Aided Translation Tools in the Software Localization Industry
115
Matthias Heyn Translation Memories: Insights and Prospects
123
Magnus Merkel Consistency and Variation in Technical Translations A Study o f Translators ’ Attitudes
137
Reinhard Schäler The Problem with Machine Translation
151
Paul Bennett Machine Translation as a Model o f Human Translation
157
Section Five: Interpreting Studies Franz Pöchhacker Unity in Diversity : The Case o f Interpreting Studies
169
Jorma Tommola and Marketta Helevä Language Direction and Source Text Complexity Effects on Trainee Performance in Simultaneous Interpreting
111
Anna-Ritta Vuorikoski User Responses to Simultaneous Interpreting
187
Contributors
195
Introduction Translation studies as a discipline has grown enormously in the last two decades. Each year sees new translation programmes, new journals, new translation titles added to the burgeoning list of publications in the area. Conferences on translation topics are organized around the globe and attract keen, committed audiences. Contributions to the discipline come from the fields of machine translation, history, literature, philosophy, linguistics, terminology, interpreting, screen translation, translation pedagogy, software localization and lexicography. This list is indicative rather than exhaustive. There is evidently great diversity in translation studies but is there much unity? Have the different branches of transla tion studies become so specialized that they can no longer talk to each other? Would translation studies be strengthened or weakened by the search for or the existence of unifying principles? It could be argued that translation studies is simply going the way of other disciplines in a world where in the sciences alone there are esti mated to be 90,000 specialisms. The path of development is the path of meiosis, with fragmentation the inevitable consequence of disciplinary expansion. Fragmentation indeed is celebrated by many postmodern crit ics as a triumph over the totalizing theories of modernity, theories that erased difference in the name of unity and sacrificed the peripheral to the centripetal coercion of ‘master’ theories. Translation, which involves lan guage and culture contact, border crossings, power brokering, is certainly an activity that is sensitive to the pressures of culture, politics and place and to the persistent danger of erasure and invisibility. The historical experience of translators has certainly been a strength in making transla tion theorists aware of how what is deemed to be marginal may in fact be central to the construction of a language, culture or society. For this very reason, it is important that in our discipline we do not practise our own forms of exclusion. In this volume we have brought together translation theoreticians from the fields of machine translation, interpreting, feminist theory, computer-assisted translation, advertising, literature, linguistics, screen translation and translation pedagogy to counter the tendency to partition or exclude in translation studies. It is important that machine translation specialists and literary translators be found between the covers of the same book if only because the nomadic journeying of concepts is often the key to intellectual discovery and renewal. Celebrating our differences does not entail denigrating the commonality of our concerns. Too often the discipline of translation studies can fetishize difference at the ex pense of memory. Linguistics was embraced with enthusiasm in the 1960s
vi
Unity in Diversity
and early 1970s and then abandoned with similar gusto in the 1980s, de spite the fact that the discipline of linguistics itself has changed radically in those years and still has much to teach us. Systems theory was champi oned and then pilloried and yet the full insights of systems theory have still to be properly applied to translation studies. Machine translation was hailed as the translation Messiah and then became the butt of a thou sand jokes repeated ad nauseam about howlers found in MT output. Yet machine translation is a very different activity from that practised in the early 1960s, a fact that is forgotten or conveniently ignored. The blind embrace of theoretical fashion can lead to much theoretical redundancy as the same observations appear endlessly in different terminological guises. Theory should be as muclj an act of memory as an act of inven tion and we should be loath to c^kst aside disciplines that are suddenly deemed to be useless or irrelevant. If it is important to remember the past, the future too should not be forgotten. New technologies, for example, are having a decisive impact on the nature of the translation profession and we must remain aware of developments that are changing and will continue to change in very real ways the working practices of translators. Harlequin, not Jerome, could in our time be considered the emblem atic figure of translation studies. Information theory, linguistics, systems theory, gender theory, semiotics, and cognitive psychology are some of the disciplines that have been appropriated by translation theoreticians to provide the perfect theoretical ‘fit’ for translation studies. Is the inter disciplinary whole greater than the sum of the disciplinary parts or do we have juxtaposition without shape, colours without form? It is striking that though we talk to each other more and more, even if frequently within rather than across translation specialisms, we seem to talk on the whole to ourselves. Countless conferences, articles, and books on history, sociology, anthropology, politics, new technologies, and phi losophy in the late twentieth century simply ignore the question of language and translation. And yet in an era of globalization, migration, mass tour ism, multilingual federal structures, postcolonial debate and exponential progress in information technology, translation studies is arguably at the very heart of our attempts to comprehend the developments in the late modem age. The articles in this volume illustrate the capacity of transla tion studies to deal with a very diverse range of phenomena, a capacity that was already evident in the Translation Studies Conference organized by Dublin City University in 1996, a conference that originally prompted the idea of a volume of articles on the theme of unity in diversity. The volume is as much about questions as answers and one recurring question is: does translation studies have a number of core distinguishing features that set it apart as a discipline in its own right?
Introduction
vii
Would greater disciplinary integration be a gain or a loss? Or is trans lation studies merely experiencing an epistemological crisis of confidence that has afflicted most areas of human enquiry since the 1960s? Transla tion Studies: Unity in Diversity? examines these questions and others in articles from many different areas of translation studies. The articles draw attention to the complex nature of the translation transaction and indeed reveal a number of cases of previously unexplored aspects of the phe nomenon of translation. The collection deliberately brings together areas of translation studies that are often addressed in separate volumes and arenas in order to suggest areas of common interest beyond the necessary boundaries of specialization. The merits of unity and the virtues of diver sity are debated from theoretical, practical and professional perspectives. Theory is rarely a stranger to orthodoxy and translation theory is no exception. The articles in the first section of this volume challenge as sumptions that are made about the nature of translation. Gender and translation, the consequences of translation for self-construction and the position of translation in the cultural and economic dynamics of globali zation are considered in addition to specific speculation on the pertinence of translation to late twentieth-century experience. Luise von Flotow, in ‘Dis-unity and Diversity: Feminist Approaches to Translation Studies, details the impressive growth in recent years of feminist scholarship in the area of translation studies. She then charts the lines of antagonism and tension that separate the different feminist theo reticians, particularly centring around the questions of history, identity and positionality. The article argues that theoretical unity is of question able value and that for feminist theoreticians of translation, diversity is not simply a consequence of the complexity of the object observed but is a desirable outcome of any theoretical endeavour. Susan Ingram exam ines the autobiographical writings of Alice Kaplan and Eva Hoffman from the standpoint of the bilingual author for whom writing is an act of trans lation. The contrasting experiences of Kaplan and Hoffman are analyzed in the context of translation as either dispossession or emancipation, a form of humiliating alienation and displacement or a powerful means of personal and cultural enrichment. Walter Benjamin’s reading of ‘the Kaf ka situation’ and the writings of Deleuze and Guattari on ‘minor’ literatures are integrated into a theory of authorial identity that is crucially depend ent on the figure of translation. Translation in national context is the theme which links the five articles by Brown and Sherlock, Sohar, St.-Pierre, Wakabayashi and O’Connell. While these authors are all writing from quite diverse cul tural perspectives, dealing with different languages (Scots, Welsh, Irish, Hungarian, Japanese and the many languages of India) and different translation processes, the question of context recurs in all of the articles.
viii
Unity in Diversity
The context or environment in which a translation is to be received will determine which translation process is most appropriate. It may be more appropriate to adapt texts to the social, cultural and political climate of the target culture, as described by Brown and Sherlock in their account of the challenges they faced in translating the ancient Greek myth of Anti gone into contemporary Scots and Welsh. Or perhaps in cases where a particular genre does not exist in a culture, it is better to create fictitious translations of works, i.e. texts masquerading as translations and which have no source text, rather than genuine translations as has happened in the case of science fiction in Hungary. The Indian experience of transla tion shows the extent to which relations of power between languages are reinforced by certain translation policy decisions. In other cases, it may be that knowledge of the source language is a prerequisite for the under standing of the target text, a notion explored in Wakabayashi’s fascinating overview of marginal forms of translation in Japan. In the case of Ireland, assumptions made about the target audience have crucial consequences for screen translation policies. Indeed, whichever process translators choose, whether to adapt, to transpose, to create or to translate in the more conventional manner, their choice is determined by their understanding of the expectations and capabilities of their audience. Brown and Sherlock discuss aspects of the nature of translation and proceed, on the basis of that discussion, to consider the nature of the Scots and Welsh languages. They discuss the nature of the process in which myth, specifically Greek myth, may be translated into another cul ture. The authors argue that the process usually called ‘adaptation’ is actually one of translation of a significant mythic structure from the pre mises of one cultural frame to another in a way analogous to the translation of text from one language to another. They also discuss the ways in which Welsh and Scots are capable of dealing with the material of the Antigone play as written by Brown within a Scots cultural frame. Finally, they raise some tentative hypotheses about modem Scots and Welsh as lan guages. Aniko Sohar examines how science fiction and fantasy novels are developed and established in Hungary through translation and pseudo translation. Paul St. Pierre draws on the plurilingual reality of modem India to argue that translation principles are not immutable, ahistorical universals but are grounded in the specific linguistic, historical and po litical experiences of a country. The notion that one should, for example, only translate into one’s mother tongue is both impractical and culturally harmful in the Indian context. He explores the translation relationships between the different Indian languages and the role of English as a filter language. St. Pierre alludes to the notion of ‘transcreation’ to demon strate how translation into English can often be a way of reinforcing rather than weakening different language identities in India. Judy Wakaba-
Introduction
ix
yashi’s article focuses on two non-prototypical methods of rendering foreign texts into a form comprehensible to Japanese readers. The first method, kambun kundoku, uses grammatical indicators and marks indi cating word order so as to allow Japanese readers direct access to Chinese texts. Another practice common throughout Japanese literary history has been adaptation, with both traditional Chinese tales and European works being adapted to varying degrees, often by famous writers who have used adaptations as a stimulus for their own creative activities. She examines the degree of acceptance in Japan of these practices as translation, and the nature of the relationship between the source and ‘target’ texts in these instances. Eithne O’Connell examines the choices and constraints that are operative in screen translation. She emphasizes the distinct na ture of dubbing and subtitling and challenges traditional explanations that are put forward to explain the choice of dubbing over subtitling. O ’Con nell stresses the importance of language politics and language planning in screen translation decisions, particularly with respect to minority lan guages. She also draws attention to factors of age, literacy levels and gender that can influence the choice of dubbing or subtitling. Theoretical work in translation studies, as in any other discipline, is powerfully informed by appropriate descriptive studies. Irena Kovacic points up the complexity of subtitling and the often significant variation in subtitling choices from one subtitler to another. Using a set of textual parameters and Halliday’s model of linguistic functions she examines the nature of differences in the work of six different subtitiers. Her article highlights the significance of language register and ideational function in the analysis of subtitles and speculates as to the impact of subtitling dif ference on audience reception. Christina Schaffner’s contribution to this volume reveals an interest in using real LI and L2 texts in discussions of translation. This time the focus is on the translation classroom, but Schaffner suggests that newly emerging text types will lead to a questioning across the board in translation theory of fundamental concepts such as parallel text and text type. Peters and Picchi describe procedures that have been developed at the Istituto di Linguistica Computazionale, Pisa, to construct and query bilingual corpora and to extract significant data for translation purposes and contrastive textual studies. They treat both parallel corpora, i.e. collections of source texts and their translations into a target language, and bilingual comparable corpora, collections of LI and L2 texts, that, although they were produced in comparable LI and L2 contexts, do not actually bear any translation relationship to one another. Sara Laviosa’s article describes the use of a monolingual English compa rable corpus to identify features that differentiate translated text from text originally written in English. She focuses in particular on the hypoth esis that simplification is characteristic of translated text. What Laviosa
Unity in Diversity and Peters and Picchi have in common is an interest in establishing rigor ous procedures for the empirical investigation of translation. Computer Aided Translation and Machine Translation have under gone considerable change over the last decade and no adequate account of contemporary research in translation studies can afford to ignore these areas. One recent development, the advent of translation memories, relies on the cooperation of humans and computers: humans do the transla tions; computers store and recycle the same translations. In her article, Sharon O’Brien explains the basic principles of translation memories and text alignment systems and their use in the software localization industry, one of the biggest users of such systems. O’Brien’s article is followed by a contribution from one of the leading industrial experts in Computer Aided Translation (CAT), Matthias Heyn. Heyn broadens the debate on translation memories with a detailed discussion of the technology involv ed, the users it serves, and its psychological and financial impact on translators. He argues that the relatively simple interfaces that translation memory (TM) systems come with belie the complex internal functioning of such systems, and he debunks the notion that it is just the software sector that uses TM systems, stressing that user profiles are changing, and broadening, all the time. The growth of application areas for this technology in turn means that the core functionality of TM systems has to be extended to cope with the diversity of user needs, and Heyn provides a much more fine-grained classification of these needs than have been seen in previous discussion. Magnus Merkel, in his contribution, offers a detailed analysis of the responses of technical translators, project leaders and translation customers to the issue of variation and consistency. He does this specifically in relation to the use of translation memories and other CAT tools. Merkel finds that translators on the whole are positive in their attitude towards translation tools and value consistency in techni cal translation. However, he finds evidence of considerable variation in what translators deem to be the ‘best’ translation of particular segments and wonders whether translators will always be willing to accept the translations suggested by TM based programs. Customers also expressed certain reservations about excessive reliance on translation memories that had not been properly verified. More generally, among technical transla tors themselves there was a strong awareness of the importance of functional contexts in dictating translation choices. Human Translation and CAT have not always been seen as compat ible, and Reinhard Schäler would argue that many human translators still have (misguided) misgivings about the role computers are playing in the translation process. Schäler argues that traditional translators have a sys tem of values and references that can make them ill-disposed towards CAT. Translation realities, however, mean that CAT is being used more
Introduction
xi
and more and translators are being excluded from significant develop ments in their own field. The article calls, among other things, for real integration of technology in translator training (as opposed to add-on modules) with a view to bringing about a change in many translators’ mindsets. A final area where there can be fruitful exchanges between CAT and Human Translation is in the theoretical arena. Paul Bennett discusses what the respective theories of Machine Translation (MT) and Human Trans lation have to learn from each other. He asks to what extent current MT systems attempt to model the process of human translation, and questions whether this would be a useful design criterion for MT systems in the first place. Interpreting studies have been the site of much research in recent years and a burgeoning research literature. However, in the extreme diversity of conflicting approaches it is not always possible to establish ways of transcending old divisions. In addition, certain elements of received wis dom in interpreter training with respect to language direction and user expectations may in fact be contradicted by the evidence. Franz Pochhacker in ‘Unity in Diversity: The Case of Interpreting Studies’ begins by examining the main paradigms in contemporary interpreting research and establishing what common ground exists among the different approaches. He then outlines ways of transcending the old antagonism between cul tural/humanist forms of enquiry and scientific methods of investigation in interpreting research. The article sees diversity not as an obstacle to re search in interpreting but as a way forward, methodological complexity and variety capturing the complex nature of the phenomenon of interpret ing itself. Jorma Tommola and Marketta Heleva note that the general practice in interpreting training is to require that students interpret only into their mother tongue. In this article, the wisdom of such practice is questioned on the basis of a study carried out on twelve trainee interpret ers in Finland. The authors argue that a central aspect of interpreting is the creation of a rich semantic representation of the source text. In going from the A language to the B language, the interpreter will experience fewer comprehension problems, therefore the performance is likely to be more fluent and accurate particularly where speeches are highly techni cal or structurally complex. The authors use a proposition-based scoring method to demonstrate the effect of complexity and language direction on interpreting quality. Anna-Ritta Vuorikoski’s main concern is the responses of customers or end-users of interpreting services. The article is based on a survey of user responses to interpreting at five different seminars where English and Finnish were the working languages. The author argues that any satisfactory approach to interpreter training has to take into account users’ perceptions where, for example, informational accuracy is often
xii
Unity in Diversity
considered to be far more important than TL fluency. The article also high lights patterns of usage so that in the case of English, the use of strong regional varieties by speakers will lead to heavy demand on interpreting services which in turn has implications for the education of interpreters. The article gives a full picture of user responses to simultaneous interpret ing in a variety of different contexts. The collection offers, we believe, a valuable overview of the current state of translation studies and contains articles that we hope will make an important and useful contribution to the development of the discipline at both a practical and theoretical level.
The Editors December 1997
Section One The Nature of Translation
Page Intentionally Left Blank
Dis-Unity and Diversity Feminist Approaches to Translation Studies LUISE VON FLOTOW
School of Translation and Interpretation University of Ottawa, Canada Feminist work in translation studies is growing increasingly di verse. This may be ascribed to the focus on difference that has developed over the past fifteen years in feminist scholarship gener ally, but it is also a result of the cross-cultural work that translation studies entails. This article discusses a number of current examples of dis-unity within feminist work in translation, and locates them in the contextual and cultural differences that obtain between the participating scholars. It then discusses the extent to which factors such as ‘identity politics ’, \positionality' and ‘historicity' have had an effect on insights and value judgements in these specific cases and in other areas o f feminist work in translation studies. The focus is thus on dis-unity, diversity and complexity, factors which appear to be leading to highly productive work. Issues of unity, on the other hand, remain problematic. Introduction1 Feminist work in translation and translation studies is diversifying; it is not only extending the bounds once posed by gender difference and confronting assumptions that derived from them, it is beginning to ex plore what theorist Alice Parker (1993) has tentatively termed poly sexual and multigendered approaches to translation. Parker writes about lesbian translation; my focus in this article is on the complexity and diversity, and on the dis-unity of the more conventional feminist work being done in the field. Nonetheless, Parker’s insights are useful.Writing from the still marginal position of lesbian cultural criticism, she emphasizes the ap preciation of diversity as a survival strategy: Survival depends on a nonreductive appreciation of diversity and complexity that cluster around two poles: responsibility (the ability to respond), and desirabilty (the ability to desire) (1993:330). Her comment indicates that still more is needed than just an apprecia tion of diversity. Criticism cast as ‘response-ability’ and ‘desire-ability’ 1I am grateful for post-doctoral funding from the Social Sciences and Human ities Council of Canada which made it possible for me to work on this text.
4
Unity in Diversity
is required for survival. Diversity and complexity develop, unfold, even mutate into some form of productive progress in response to responsive and desiring criticism. Given the current level of criticism which I will focus on in this article, it seems that mainstream feminist approaches to translation and translation studies have moved well beyond concerns about survival. If diversity and complexity in any academic discourse are meas ures of the interest it awakes, then the response that feminist work in translation is soliciting is encouraging. Lack of critical response is after all dead silence, w hether it be deliberate as in the German term totschweigen (to silence to death), or simply due to lack of interest or ap athy. But univocity and consensus also shut down development. Luckily, perhaps, critical responses in regard to feminist work are rarely neutral; factors of cultural or ideological conditioning, academic ambition, or insti tutional constraints are inevitably involved. Not much danger of consensus. My focus in this article will thus be on two aspects of feminist work in translation: its current diversity and dis-unity, and the factors underlying this state of affairs, and indeed much contemporary work in feminist criti cism. Broadly speaking, the diversity in feminist translation studies is one desirable result of changes that have taken place in feminist thought over the last twenty years. These changes moved essentialist feminist thinking that once viewed all women as sharing more or less similar forms of so cial, cultural, economic and political oppression to more differentiated approaches in which cultural, ethnic, economic, and many other differ ences between women are recognized and brought to bear in critical discourses. Given the cross-cultural work that translation entails, nonreductive differentiation is doubly present in feminist approaches to translation studies - between women and between cultures. This focus on difference is also strategic, as Linda Alcoff (1988/1994) has argued. She advocates that feminist thinkers acknowledge three factors in their work in order to avoid gross (essentialist) generalizations and the dissemin ation of culturally and politically questionable material about women or feminisms, and in order to thus negotiate the difficult ideological and cultural rifts that divide women. These factors are: identity politics (the writer/critic’s identity has an effect on their perceptions and writing), positionality (the effect of this identity is relativized by institutional, economic, and other factors) and the historical dimension (perceptions/ interests/topics change with the times as does identity). I will discuss a number of examples to show the role these factors play in the current diversity in feminist work in translation studies. Disunity in feminist work: undermining consensus Disunity in feminist approaches to translation and translation studies has
von Flotow: Dis-unity and Diversity
5
recently become visible in a number of different publications. The issues I will discuss coalesce as: (shoddy) mainstream English translations of third world women’s texts for anglophone consumption; elitist and inac cessible work which has little to do with the socio-political concerns often ascribed to Anglo-American feminisms; theoretical incoherence and hy pocrisy in feminist translation and feminist critique of patriarchal theories. Mainstream ‘translatese ’ o f third world material In a text on the politics of translation, Gayatri Spivak discusses the trans lation into English and French of writing by third world women. Among the issues she raises is the problem of ‘with-it translatese’ (1992:180) which, she says, serves to construct a largely misrepresentative view of third world women’s texts. The implementation of English or French ‘translatese’, she says, disregards the rhetoricity of the source text and focuses on making as much women’s writing as possible available to the West/North. Ironically, the resulting misrepresentations are due to work by mainly Anglo-American women who, in gestures of feminist solidar ity, want to make third world women’s writing available in English. However, their ‘translatese’, the language of mainstream anglophone translation, obscures the differences between women of very different and differently empowered cultures, ostensibly in order to make the texts ‘accessible’. Further, it deprives the texts of their individual styles, styles which are by no means homogeneous within one particular culture, let alone across third world writing. As Spivak puts it: ‘the literature of a woman in Palestine begins to resemble, in the feel of its prose, something by a man in Taiwan’ (ibid). For Spivak, these gestures of apparent femi nist goodwill are in effect applications of the ‘law of the strongest’, which, at this particular moment, endorses translation into English as the easiest way of being ‘democratic with minorities’. Thus, what may have started out as feminist attempts to understand and make available third world women’s experiences and writing turns out, in Spivak’s estimation, to be appropriation, misrepresentation and the salving of guilty consciences. Elitist translation A recent text by Canadian Robyn Gillam (1995) criticizes feminist ap proaches to translation from a different angle, this time within the field of Canadian ‘feminist iconography’. She compares a number of English translations of work by Quebec radical feminist Nicole Brossard, and suggests that translations produced from a deliberately feminist per spective (i.e. translations of Amantes or L ’Am èr by Barbara Godard) make the already difficult source material even more obscure by produc ing English texts that privilege sound associations and extend already
6
Unity in Diversity
complex wordplay. Further, in order to produce these effects in English, the translator has deliberately mistranslated. Gillam suggests that these translations are addressed to a small academic elite that is already bilin gual and can at most marvel at the linguistic virtuosity of both author and translator. Her criticism is based on the view that French-speaking and English-speaking Canadians have inherently different political relation ships to their respective languages. For the Québécois, language has been and continues to be a political issue of daily life, while this tends not to be the case for most other Canadians. Thus, the distortion/deconstruction of language in itself means something different in a text in Quebec than in English-speaking Canada. It has a different political value, because of the different cultural history inscribed in it and by it. Brossard’s decon struction of patriarchal language takes this a step further, a step that ‘works’ for Québécois readers since they are more sensitized to political use of language. For English-speaking Canadians her writing may be little more than exotic, however, and its reduction through translation to “an intellectual game where there exists nothing but words and their mean ings” (1995:11) hardly propagates what Gillam views as the socio-political goals of feminist writing and work. This critique has, of course, challenged the response-ability of Barbara Godard, whose translations are under attack. In her reply, she stresses the context of the early 1980s in which her translations were done: a time of burgeoning feminist activity, of film-making and journal-founding, a time that welcomed and celebrated feminist creativity, and thus authorized her version of feminist translation (Godard 1996). But she also responds to the ‘separatism and classificatory demarcation’ that she claims is a corol lary to Gillam’s text, where anglophone social activism is contrasted to francophone ‘epistemological and cultural revolution’ (1996: 40). Trans lation, for Godard, strives to traverse precisely these types of boundaries. Hypocritical translation Another critique of the feminist approach to translation praxis and theory advocated by Godard and numerous others has appeared in recent work by Rosemary Arrojo (1994, 1995). Arrojo takes issue with what she has called the ‘hypocritical’ (1994:160), ‘anxious’ (ibid) and ‘theoretically [not] coherent’ (ibid: 149) work by predominantly anglophone women and men who apply feminist activism to translation. To summarize very briefly, these translators (Suzanne Levine, Carol Maier, Barbara Godard, Susanne de Lotbinière-Harwood, Howard Scott, among others) discuss and sometimes assume the right to intervene on a political level in the text they are translating. They do so in a number of places: for example, where they consider it necessary to mitigate ‘offensive’ forms of ma chismo or misogyny, where they consider it appropriate to make explicit
von Flotow: Dis-unity and Diversity
7
the feminist message or rhetoricity that may be implicit in the source text, where they want to implant feminist thought that does not necessarily ex ist in the source text. Arrojo bases her comments on three points: for one, she views some feminist translators’ claims that their work is faithful to the tenor of the source text as not congruent with their openly feminist politics, and therefore theoretically incoherent as well as anxious: they cannot let go of the ‘fidelity ethic’, though they deliberately undermine it. For another, she views feminist criticism of ‘male violence’ in transla tion and translation theory as no less violent, and therefore hypocritical. Thirdly, she considers the generalized references to post-structuralist theories with which some textual interventions are justified travesties of these theories. What she perhaps does not see is the historical dimension of these approaches; feminist translation of the 1980s develops post-essentialist insights about women, opening up views of women as nurturing peace-lovers and, instead, exploring women’s aggression, anger and agency. (A good example of the literary expression of this anger and viol ence are images evoked in Brossard’s L ’Amer: “I have killed the womb and am writing it”, or her notion of recycling the uterus as a bookbag). Feminist translators’ ‘violence’ may well be an extension of this type of discourse. Arrojo may also have overestimated the limited and highly focused nature of feminist interventions in translation; their political rheto ric sometimes does outstrip the actual interventions carried out on translations. While Arrojo rightly sees that these feminist translators and academics are as politically invested and biased as the authors or the theorists they seek to undermine - a bias they do not conceal -, she does not acknowledge her own political investment, her own ‘positionality’. What is one to make, for instance, of the following concluding statement: After all, if we cannot be really faithful to the texts we translate, if we cannot avoid being faithful to our own circumstances and per spective, we should simply make an effort to accept and be open about our ‘infidelities’ and try to forget the unnecessary guilt they bring (1994:160). Apart from wondering about the irritation these rather sanctimoni ous sentiments might arouse, I am led to ask who ‘we’ are, and exactly how ‘we’ should go about ‘simply’ accepting our ‘infidelities’. Should all translators in all cultures regardless of historical, cultural and economic inequalities or differences strive for this state of grace? Should all aca demics involved in translation studies do so? Or does this address only the women involved in some way with translation? If so, then given the differences between women and women’s cultures that Gillam and Spivak have most recently addressed, and that Arrojo has amply demonstrated, this may be a vain exhortation.
8
Unity in Diversity
To briefly comment on these examples of dis-unity: Gillam’s criti cism seems to be diametrically opposed to Spivak’s; she wants the feminist text made meaningful and accessible for the translating culture and its feminist activists of all kinds, while Spivak calls for a translation practice that resists the homogenizing demands for ‘easy-reading’ of the target culture feminist reader. The gulf between these two positions may be un derstood as a difference in perspective, a combination of the three factors that Alcoff has delineated. Gillam is an anglophone Canadian academic concerned with two white middle-class approaches to writing in Canada, writing in languages and cultures that exist side by side within one politi cal entity at the same historical moment. She expects feminist interaction on more than just an academic level, and wants it moved out of academia and popularized. Her critique is local, and motivated in part by an irrita tion with the ‘consensual’ aspects of certain feminist discourses in Canada which threaten to disable discussion (private correpondence). Spivak, on the other hand, is concerned with languages and cultures, whose rela tionship is marked by glaring economic inequalities and a history of colonization. In such a context, she sees translation that popularizes wom en’s work by producing accessible texts as another form of imperialism, or worse as a way to salve the consciences of the more privileged West ern feminists, while allowing them to further their careers. One wonders whether Spivak’s own conscience at her privileged diasporic position is also salved by this critique of mainstream anglophone translation. Arrojo, on the other hand, objects to the confrontational approach adopted by some Anglo-American feminists and justified with reference to their readings of post-structuralist theories. It is debatable whether the contradictions and ‘hypocrisy’ Arrojo dwells on are due to mis-readings, as she claims and which she seeks to correct, or whether these are simply other ‘strate gic’ readings. (Spivak has in the past advocated women’s strategic use of what they find useful in the brew of post-structuralist theories). To conclude this dis-unity section: the diversification of feminist dis courses on the subject of translation is a noticeable recent development, visible not only in these critical writings but also in numerous papers and publications, which will hopefully lead to further dis-unity. One rather unfortunate unifying aspect does seem to exist in all the diversity, how ever: feminist work is largely produced by anglophones or in response to translations into English. This is as much the case for recent comparative studies on new Sappho and Louise Labe translations (Batchelor 1995; Prins 1996) as it is for theories of lesbian translation (Parker 1993; Marlatt 1989) or feminist rewritings of the Bible (Haugerud 1977; An Inclusive Language Lectionary 1983).2 Similarly, recent work on women translators of the Renaissance and the nineteenth century which scruti nizes their writing strategies and ‘positionality’ is in English and focused
von Floîow: Dis-unity and Diversity
9
on English materials (Robinson 1995; Krontiris 1992; Zwarg 1990). The same applies to a text that examines ‘marginal’ Irish women’s writing in translation (O’Connell 1995), an article that presents the multiplicity of the Pandora figure as a feminist response to the search for linguistic Babel ian unity (Littau 1995) and my own work on translations of experimental feminist writing (von Flotow 1991; 1995; 1997). This predominance of anglophone work in feminist approaches to translation studies brings me to the three factors delineated earlier that affect feminist research, mak ing it possible and highly productive on the one hand, and dis-unifying it on the other. Factors motivating 6responsible and desirable’ disunity Contemporary feminist scholarship in English purposely stresses the rela tive nature of its approach. A recent introductory statement by Nicole Jouve Ward (1991 :vii) is not untypical: No one who writes today can or should forget their race and their gender. The ‘I’ who has written this book is white: privileged, yes, middle class, yes; and everything it has to say is limited and col oured by unconscious western European assumptions. W ard’s ‘autobiographical’ approach to criticism is carried through large sections of her work, demonstrating the extent to which a critic’s subjectivity and position as well as the contingent nature of her particular historical context constitute her perceptions and her approach. Like Ward, writers such as Spivak, Parker, and de Lotbinière-Harwood are aware of the relative nature of their insights, and of the relative value of these, coming as they do from specific individuals at specific historical mo ments. For translation studies, such applications of ‘identity politics’, ‘positionality’ and the references to ‘historical context’ are important in struments with which to get at and explore differences lodged in cultural knowledge and conditioning as well as in the various institutional factors that hamper or enhance women’s participation in academic and transla tion work. They also play some role in the dis-unity of feminist work in translation studies and in the reasons for its largely anglophone bias. In the balance of this article I will discuss these terms, and give examples of their effect in the field. ‘Identity politics’ acknowledges the academic’s personal interests and 2 Some work is being done in German and Dutch: cf. an unpublished paper by Dr. H. Wegener, ‘Frauengerechte Sprache in der Bibel in Blick auf die Revision der Guten Nachricht’ cited in Ellingworth (1992) and A. de Vriess ‘Sexism in Modern Bible Translation’ (paper given at EST Congress, Prague, 1995).
Unity in Diversity
10 needs, and as Alcoff explains, is based on
the initial premise that all persons, including the theorist, have a fleshy, material identity that will influence and pass judgement on all political claims ... [It] ... introduces identity as a factor in any political analysis (1988/1994:116). In feminist approaches to translation studies, identity politics thus in corporates the academic’s identity as a specific individual, with certain consciously identifiable cultural/political characteristics that will deter mine her or his insights, opinions and prejudices. Texts by Gayatri Spivak (1988; 1992) are marked by such identifications: she writes in the first person singular - there is no assumed consensual ‘we’ -; she describes how her thinking has developed from conversations, from her own trans lations of Bengali and French material, from her experience of Western feminisms and ‘Indian’ women’s issues, as well as from the experience of the Indian caste system and the American academic environment. All these factors affect how she translates, and how she thinks about transla tion, and although her political stance is not always clearly delineated, it is placed into context. Writing on translation by Susanne de LotbinièreHarwood (1995) takes a very similar approach. ‘Positionality’, according to Alcoff, further relativizes the situation by making identity relative to a constantly shifting context, to a situation that includes a network of elements involving others, the objective economic conditions, cultural and political institutions and ideologies, and so on (ibid: 116). This concept not only allows researchers/academics to acknowledge and account for constantly shifting personal and intellectual settings and the effects of such shifts on scholarly ‘knowledge’ and analyses, but it can also be used as a fluid location from which to construct meaning, a per spective from which values are interpreted and constructed - differently at different times. Examples of how this has affected translation studies can be found in the burgeoning feminist work during the 1980s, as women translators and academics began to apply to translation what they were learning and experiencing as a result of the women’s movement and femi nist enquiry in the universities. Texts by Suzanne Levine, Carol Maier and Marlene Wildemann have clearly developed out of the changing fo cus that comes with a change in positionality. At least in the anglophone countries of North America and Europe, networks, economic conditions and cultural institutions have been such that values could be reinterpreted and newly constructed under feminist ‘pressure’. The ‘positionality’ fac tor may also aid in understanding the apparent dearth of fem inist
von Flotow: Dis-unity and Diversity
11
scholarship in translation studies outside the English-speaking commu nity; (I am aware that the term ‘dearth’ implies a value judgement that stems from my own positionality). An interesting question in this regard might focus on the institutional and economic factors that are involved when hundreds of works of both mainstream and experimental feminist writing are translated from English into various European languages, yet virtually no theoretical or analytical work has been stimulated by this massive influx of translations. The third factor - ‘the historical dimension’ of scholarly discourse - is used by Alcoff to articulate a concept of gendered subjectivity “without pinning it down one way or the other for all time” (ibid: 114), construing it instead “in relation to concrete habits, practices and discourses, while at the same time recognizing the fluidity of these” (ibid: 115). For Alcoff the ‘historical dimension’ is the factor that makes gendered subjectivity change with the times and with the political and institutional constella tions that determine concrete options, real possibilities, or real obstacles. It is a powerful factor in both identity politics and positionality. This factor has been seen at work in Gillam’s critique of Godard, and in Godard’s response, as well as in Spivak’s post-colonial criticisms of mainstream translation. It is also at work in recent material on metaphors of translation. In one of the first books to suggest that translation is a form of ‘manipulation’, Theo Hermans (1985) discusses the changing metaphors used to describe translation in the Renaissance. On two occa sions Hermans refers to the Earl of Roscommon’s Essay on Translated Verse, and the much-cited admonition that the translator ... chuse an Author as you chuse a Friend: United by this Sympathetick Bond, You grow familiar, intimate and fond For Hermans, this advice by Roscommon is a sign of the Renaissance translator’s developing sense of a personal relationship and affinity with the author which leads to a new approach to translation. Only three years later an essay by American scholar Lori Chamberlain (1988/1992) uses the same passage by the Earl of Roscommon for a rather different pur pose. Chamberlain’s focus is feminist, a ‘positionality’ that may be due to her location within 1980s North American academia, and she ‘sees’ that Roscommon’s exhortation to the translator stems from patriarchal ideology. For her, the language used by Roscommon’s author/translator “echoes that of conduct books and reflects attitudes about the proper dif ferences in educating males and females” (ibid:59); and the ‘Sympathetick Bond’ between author and translator is a specific form of male bonding in the “struggles for the right of paternity” (ibid:58). Doubtless, Cham berlain’s feminist slant on the Earl of Roscommon is as much an effect of
12
Unity in Diversity
the ‘historical dimension’ of scholarship as is Hermans’ failure to ad dress feminist issues in 1985. In brief conclusion, it would seem that feminist work in translation continues to expand and develop in the 1990s, fuelled on the one hand by institutionally sanctioned interests in gender difference in some parts of the world, and on the other by conflicts between scholars that may stem from cultural, ethnic, ideological and institutional affiliations. The com plexities and dis-unities resulting from the interplay of these factors are more productive, however, than consensus on the sometimes sensitive issues they address. References Alcoff, L. (1988/1994) ‘Cultural Feminism versus Post-Structuralism: The Identity Crisis in Feminist Theory’, in Nicholas Dirks, Geoff Eley and Sherry Ortner (eds) A Reader in Contemporary Social Theory, Princeton, N.J.: Princeton University Press, 96-122. An Inclusive Language Lectionary (1983) Atlanta: John Knox Press, New York: The Pilgrim Press, Philadelphia: The Westminster Press. Arrojo, R. (1994) ‘Fidelity and the Gendered Translation’, TTR. Traduction Terminologie Rédaction, VIII(2): 147-64. Arrojo, R. (1995) ‘Feminist ‘Orgasmic’ Theories of Translation and their Contradictions’, TradTerm 2: 67-75. Batchelor, J. (1995) ‘Changing the Agenda: Gender Consciousness in Relation to Louise Labé’s Sonnets’. Paper presented at the EST conference, Prague. Brossard, N. (1977) UAmèr ou le chapitre effrité, Montréal: Quinze; trans. Godard, B. (1983), as These Our Mothers, Toronto: Coach House Press. Brossard, N. (1980) Amantes, Montréal: Quinze; trans. B.Godard, 1987, as Lovers, Montreal: Guernica Editions. Chamberlain, L. (1988/1992)~‘Gender and the Metaphorics of Translation’, in L. Venuti (ed) Rethinking Translation. Discourse, Subjectivity, Ideol ogy, London & New York: Routledge, 57-74. De Lotbinière-Harwood, S. (1995) ‘Geographies of Why’, in S. Simon (ed) Culture in Transit. Translating the Literature o f Quebec, Montreal: Véhicule Press, 55-68. Ellingworth, P. (1992) ‘The Scope of Inclusive Language’, The Bible Trans lator 43(1): 130-140. Gillam, R. (1995) ‘The Mauve File Folder. Notes on the Translation of Nicole Brossard’, Paragraph 17(2): 8-12. Godard, B. (1996) ‘Negotiating Relations’, Paragraph 18(1): 39-40. Haugerud, J. (1977) ‘Introduction’ to The Word for Us, Seattle, WA.: Coa lition on Women and Religion. Hermans, T. (1985) ‘Images of Translation. Metaphor and Imagery in the Renaissance Discourse on Translation’, in T. Hermans (ed) The Manipu lation of Literature. Studies in Literary Translation, London & Sydney:
von Flotow: Dis-unity and Diversity
13
Croom Helm, 103-135. Jouve Ward, N. (1991) White Woman Speaks with Forked Tongue, London & New York: Routledge. Krontiris, T. (1992) Oppositional Voices: Women as Writers and Transla tors of Literature in the English Renaissance, London & New York: Routledge. Littau, K. (1995) ‘Pandora’s Tongues’. Paper given at EST conference, Prague. Marlatt, D. (1989) ‘Translating MAUVE: Reading Writing’, Tessera VI: 27-30. O’Connell, E. (1995) ‘Twice Marginalized: The Translation of Contempo rary Irish Women’s Poetry’. Paper given at EST conference, Prague. Parker, A. (1993) ‘Under the Covers: A Synaesthesia of Desire. Lesbian Translations’, in Susan J. Wolfe and Julia Penelope (eds) Sexual Prac tice, Textual Theory: Lesbian Cultural Criticism, Cambridge & Oxford: Blackwell, 322-339. Prins, Y. (1996) ‘Sappho’s Afterlife in Translation’, in Ellen Greene (ed) Rereading Sappho: A Collection of Critical Essays, Berkeley: University of California Press, 36-67. Robinson, D. (1995) ‘Theorizing Translation in a Woman’s Voice, Subvert ing the Rhetoric of Patronage, Courtly Love and Morality’, The Translator 1(2): 153-175. Spivak, G. C. (1988) In Other Worlds: Essays in Cultural Politics, New York & London: Routledge. Spivak, G. C. (1992) ‘The Politics of Translation’, in Michèle Barrett and Anne Phillips (eds) Destabilizing Theory, Stanford: Stanford University Press, 177-200. von Flotow, L. (1991) ‘Feminist Translation: Contexts, Practices, Theories’, TTR IV(2): 69-84. von Flotow, L. (1995) ‘Translating Women of the Eighties: Eroticism, An ger, Ethnicity’, in S. Simon (ed) Culture in Transit. Translating the Literature of Quebec, Montreal: Véhicule Press, 31-46. von Flotow, L. (1997) ‘Mutual Punishment? Translating Feminist Word play: Mary Daly in German’, in Dirk Delabatista (ed) Traductio: Essays on Punning and Translation, Manchester: St. Jerome Press & Namur: Presses Universitaires de Namur. Zwarg, C. (1990) ‘Feminism in Translation: Margaret Fuller’s Tasso’, Stud ies in Romanticism 29: 463-490.
Page Intentionally Left Blank
Translation, Autobiography, Bilingualism SU SA N INGRAM 1
Comparative Studies, University o f Alberta, Canada For Walter Benjamin, the situation of Franz Kafka involved en countering the fragments of one's own existence and uniting their diversity into a whole within the context of a role. How can Ben jamin ’s analysis be applied to the process of autobiographical writing and translating? This article will explore the consequences of examining the theme o f ‘unity in diversity ’ at the individual level of the bilingual author whose writing is an act of translation. The recent autobiographical writings of Alice Kaplan and Eva Hoffman are exemplary in what they convey about the construc tion of authorial identity between languages and in how they model the productivity of Deleuze and GuattarVs minor literature. It is not unusual for writers of literary autobiographies to thematize lan guage and the writing process, which is, after all, an integral part of their lives. Nor is it unexpected to find in these postmodern, post-colonial times that the concept of liminality, of ‘betweenness’, has come to hang “like a banner over much recent autobiographical writing” (Hokenson 1995:95). That autobiography has become “a favorite province of “minority” lit erature” (Stelzig 1992:9), minority meaning marginal, non-‘heterosexual WASP male’, reflects the fact that writers are increasingly experiencing living and narrating that lived experience as an act of translation. It is precisely because recent theory has redefined the self in these marginal terms “as a position, a locus where discourses intersect” (Nussbaum 1988:132) that one would expect the studies of autobiography and trans lation to converge, and converge they do in Eva Hoffman’s (1989) Lost in Translation: A Life in a New Language and Alice Kaplan’s (1993) French Lessons: A Memoir. As their titles suggest, the learning of lan guage and the act of translating determine the context around which both have chosen to structure their autobiographies. However, their approaches are markedly different. Hoffman’s narrative strives for unity, her project is one of: translating backward. The way to jump over my Great Divide is to crawl backward over it in English. It’s only when I retell my whole story, back to the beginning, and from the beginning onward, in one language, that I can reconcile the voices within me with each other;
1Research for this article was supported by SSHRC.
16
Unity in Diversity it is only then that the person who judges the voices and tells the stories begins to emerge (1989:271-2).
Whereas Hoffman concentrates on reconciliation, Kaplan writes a Canetti-esque ode to diversity, the diversity she attributes to having ac quired a second language: “I’m grateful to French ... for teaching me that there is more than one way to speak” (1993:216).2 While these texts tend in opposite directions, the end-point of their narrative trajectories is very similar. Hoffman’s description of herself at the end of Lost in Trans lation, “I know that I’m a recognizable example of a species: a professional ... woman ... - one of a new breed, born of the jet age and the counter culture and middle-class ambitions and American grit” (ibid: 170), is equally true of Kaplan. Both write their autobiographies as East-Coast intellectuals with PhDs in literature - Hoffman’s from Harvard, Kaplan’s from Yale. Both construct self-narratives out of increasingly common “collisions of a dual heritage” (Hokenson 1995:104), and these narra tives self-reflexively explore the extent to which they see themselves as constructed through language and translation. Approaching Kaplan’s and Hoffman’s works as Kafkaesque encounters with learning, the first sec tion of this paper will examine these textual selves and their movement along the continuum of source and target language, mother and other tongue, while the second section will connect this movement to readings of Kafka and minor literature, with the goal of determining the extent to which the Kafka factor is useful in understanding the dynamics of unity and diversity and how they drive the translating process. How then can a moment of non-recognition motivate and structure the bilingual author of autobiography’s encounter with learning? Writ ing as one incompletely and yet inextricably assimilated into American culture, Eva Hoffman draws both thematic and stylistic attention to the moment at which she felt herself thrust into the Kafka situation. She makes explicit that it was her family’s emigration from Cracow to Vancouver when she was thirteen that set this process into motion, first by dividing the book into sections on Paradise, which recounts her childhood years 2 Besides the one overt reference to Canetti’s experience learning German from his mother (1993:129), his influence is palpable. That Kaplan thematizes the infiltration of French into her memories (ibid:214) calls to mind the way in which Canetti describes his memories as having translated themselves into German: ‘Alle Ereignisse jener ersten Jahre spielten sich auf spanisch oder bulgarisch ab. Sie haben sich mir später zum größten Teil ins Deutsche übersetzt’ (1977:15). Her parents speak French so that she and her brother won’t understand (ibid:206) in the same way that Canetti’s do German (ibid: 15). As well, the ‘founding im age’ of her work, her father in headphones (ibid: 197), is eerily similar to the imagery of the second volume of Canetti’s autobiography, Die Fackel im Ohr.
Ingram: Translation, Autobiography, Bilingualism
17
in Cracow; Exiley which details her reactions to life and adolescence in Vancouver; and The New World, where she comes of age studying Eng lish at Rice and Harvard and then working as a writer and editor in New York (and it is no accident that the book ends in the harmonious setting of a garden). Also, as reviewers such as Raimund (1994) have noted, she imbues the first section on her childhood memories of post-war Cracow with a pre-lapsarian grace that is absent in the final two sections. Whereas the opening section radiates wholeness and harmony, her existence in Vancouver is portrayed as profoundly alienated. Her own movements become strange to her: I feel less agile and self-confident with every transformation. I hold my head rigidly, so that my precarious bouffant doesn’t fall down, and I smile often, the way I see other girls do, though I’m careful not to open my lips too wide or bite them, so my lipstick won’t get smudged. I don’t know how to move easily in the high-heeled shoes somebody gave me (1989:109). Her own voice becomes strange: My voice is doing funny things. It does not seem to emerge from the same parts of my body as before. It comes out from somewhere in my throat, tight, thin, and mat - a voice without the modulations, dips and rises that it had before, when it went from my stomach all the way through my head (ibid: 121-2). Even her own name becomes strange to her, no longer her own: My sister and I hang our heads wordlessly under this careless bap tism. The teacher then introduces us to the class, mispronouncing our last name - “Wydra” - in a way we’ve never heard before ... Our Polish names didn’t refer to us; they were as surely us as our eyes or hands. These new appellations, which we ourselves can’t yet pronounce, are not us. They are identification tags, disembod ied signs pointing to objects that happen to be my sister and myself (ibid: 105). This sense of trauma is not to be found in Kaplan’s experience with French, on the contrary. In recounting her baptism into French, an event which took place, as it did for Hoffman, during adolescence upon begin ning to attend school in the other language, she writes simply: We sat in the living room for mail call and listened for our names to be read, in French, which was how I first heard my name pronounced French style with the accent on the second syllable, ah-LEASE.
18
Unity in Diversity We had our first study hall before breakfast, our second before din ner, then we walked across a lawn to the dining hall in the old building where the boys lived (1993:52).
Kaplan does not see her name becoming strange so much as foreign. It is still her name, just varied in an exotic, sophisticated manner to match her new exotic, sophisticated environment. That she herself aspires to this change is demonstrated in her answer to the question “why do people want to adopt another culture? Because there’s something in their own they don’t like, that doesn’t name them” (ibid:209). She prefers her French name and speaks with pride of being nicknamed by the first students she takes to Paris: “It was good to be alive with them and answer to the name they gave me” (ibid:208). Alienated, then, not from but in her mother ton gue, French represents for Kaplan escape, freedom, salvation and maturity: speaking a foreign language is ... a liberation from the ugliness of our received ideas and mentalities (211). At home I was the worst in sports; here [in Switzerland at boardingschool], miraculously, I was good. It felt like my life had been given to me to start over. French had saved me (57). Learning French and learning to think, learning to desire, is all mixed up in my head, until I can’t tell the difference ... French got me away from my family and taught me how to talk. Made me an adult (140-1). What is important about Kaplan’s Swiss-French boarding school ex perience is that it gives her a sense of personal agency that was previously missing. The awkwardness and lack of control that Hoffman experiences over her Canadian body and voice contrasts with Kaplan’s mastery of the French R and her grand régime of bread crusts and jam. Her choice to substitute French verbs for chocolate makes her feel “hard and controlled and patterned” (ibid:53); learning French as a leçon de choses allows her to control the world by objectifying and naming it: “There was the world out there, the world of Switzerland and French language, and I drew its contours and labeled it” (ibid:56). Thus, the alienation that Hoffman ex periences in having to translate herself stands in direct contrast to the self-empowerment which the opportunity to translate herself imparts to Kaplan. The Kafka situation with its emphasis on the disembodied projection of body and voice is an apt fulcrum around which to explore these two texts as it draws attention to the type of alienation present in the texts; it also introduces Kafka into the autobiographical translation equation
Ingram: Translation, Autobiography, Bilingualism
19
allowing us to focus more closely on theorizing the conflicts and para doxes in both the autobiographical and the translating processes which render them theoretically impossible tasks, at least in the eyes of someone like DeMan (Kaplan 1993:173). How, then, to read these autobiogra phies through Kafka in such a way that is informative for translation studies? I offer here two possibilities - first briefly through Benjamin (1968) and then through Deleuze and Guattari (1986). Benjamin’s essay ‘Franz Kafka: On the Tenth Anniversary of His Death’ begins with the anecdote in which a deeply depressed Potemkin signs the court documents which an unimportant little clerk named Shuvalkin naively yet brazenly forces upon him, and signs them in Shuvalkin’s own name, thereby creating chaos among the high officials. This little clerk is likened by Benjamin to Kafka’s K., but is he not rather one of Kafka’s unfinished assistants, “neither members of, nor strangers to, any of the other groups of figures, but, rather, messengers from one to the other” (1968:117)? And is this not, as well, the situation of the autobio graphical waiter, who marches in on her past demanding its signature only to find it depressed, in semi-darkness, biting its nails, to receive a vacant glance and one’s own signature back? And it is also, I would suggest, a situation which throws light on the great conflict in translation between readers quite content to fetishize and leave invisible the translator, a la Proust’s grandmother, and the desires of translators and theoreticians of translation to draw attention to the enterprise (May 1994:38; Bassnett and Lefevere 1990:2). Is it not that both aspire to Shuvalkin’s position the translator feeling herself, as the one who ventured into the cave, wor thy of having her name on the document despite the chaos that this perhaps would cause among the high officials, and the reader not willing to grant her this privilege because she feels that having read the text entitles her to those similarly subversive honours? The difference for Kafka between K. and the assistants, as related in a conversation with Max Brod, is one of hope: “I remember”, Brod writes, “a conversation with Kafka which be gan with present-day Europe and the decline of the human race. ‘We are nihilistic thoughts, suicidal thoughts that come into God’s head’, Kafka said. This reminded me at first of the Gnostic view of life: God as the evil demiurge, the world as his Fall.‘Oh no’, said Kafka, ‘our world is only a bad mood of God, a bad day of his’. ‘Then there is hope outside this manifestation of the world that we know’. He smiled.‘Oh, plenty of hope, an infinite amount of hope but not for us’”. These words provide a bridge to those extremely strange figures in Kafka, the only ones who have escaped from the family circle and for whom there may be hope (Benjamin 1968:116).
20
Unity in Diversity
It is the mechanics of this escape, this hope, that emerge in Deleuze and Guattari. Following Kafka’s own lead in introducing the category of mi nor literature, Deleuze and Guattari approach his work as a machinic assemblage, a rhizomic burrow with multiple roots and “multiple entran ces whose rules of usage and whose locations aren’t very well known” (1986:3). Minor does not designate a specific literature as much as a use to which any language may be put: “the revolutionary conditions for every literature within the heart of what is called great (or established) literature” (ibid:18).To make minor use of a language, as Kafka does, is to make use of its polylingualism, to oppose the oppressed quality of this language to its oppressive quality (ibid:26-7). One must become “a sort of stranger within [one’s] own language” (ibid:26, italics in original) to recognize what that language can and cannot do. For Kaplan, it is her year away at Swiss boarding-school that allows her to hear her mother tongue in the same way Hoffman does upon arrival in suburban Vancou ver. Just as Hoffman’s initial reaction to English is not particularly positive, “I can’t imagine wanting to talk their harsh-sounding language” (1989:105), Kaplan’s newly-trained French ears do not like what they rediscover upon returning to the Mid-west where “people’s voices sounded stretched and whiny - because of the diphthongs, I suppose. There was no control, no rhythm in the language I heard” (1993:70). However, this harsh, stretched, whiny language is the vehicle in which they choose to convey their lifestories as well as the force which gives them shape. Hoffman is confronted with “language tom from sense” (Deleuze and Guattari 1986:21): My mother and I met a Canadian family who live down the block today ... Now my mind gropes for some description of them, but nothing fits. They’re a different species from anyone I’ve met in Poland, and Polish words slip off of them without sticking. English words don’t hook on to anything. I try, deliberately, to come up with a few. Are these people pleasant or dull? Kindly or silly? The words float in an uncertain space. They come up from a part of my brain in which labels may be manufactured but which has no con nection to my instincts, quick reactions, knowledge (1989:108). And this confrontation with denuded language motivates the book’s clos ing episode of Edenic bliss: I look at the flowers; some of them I’ve never seen before; some names I’ve read but haven’t put together with the flowers them selves ... “Azalea”, I repeat. “Forsythia, delphinium”. The names are beautiful, and they fit the flowers perfectly. They are the flow ers, these particular flowers in this Cambridge garden (1989:280, italics added).
Ingram: Translation, Autobiography, Bilingualism
21
Kaplan’s book, on the other hand, begins by rendering asunder the precarious unity of Hoffman’s closure. “Everythingilikeiseitherillegalimmoralorfattening” - Kaplan learns this phrase by repeating it again and again until it makes her dizzy and all the words dissolve into one, and is charmed by the fact that it works, it makes her family laugh, even though she herself has no idea what it means (1993:4). That this deterritorializing of sense, this gypsy-like robbing the cradle (Deleuze and Guattari 1986:17), goes against our proclivities is thematized by both Kaplan and Hoffman in their explicit association of language with Heimat. According to Kaplan, language teachers, like analysts, “are always in search of the foolproof method that will work for any living language and will make people perfectly at home in their acquired tongue” (1993:130), while Hoffman concludes that, “it’s not that we all want to speak the King’s English, but whether we speak Appalachian or Harlem English, or Cockney, or Jamaican Creole, we want to be at home in our tongue” (1989:124). But how is one to be at home in one’s language and yet still maintain, as does Kafka’s Great Swimmer, that “I have to well admit that I am in my own country and that, in spite of all my efforts, I don’t under stand a word of the language that you are speaking” (Deleuze and Guattari 1986:94)? One answer, as we have seen, is Shuvalkin’s. For Deleuze and Guattari, however, this is, in fact, a non-question because, for them, Kafka and his assistants are part of the rhizomic paradigm of a burrow in which all of the entrances offer equal opportunity of escape and equal possibility of cave-in, blockage. Desiring escape out of all entrances at once and yet meeting blockage after cave-in at every turn marks what Reda Bensmaia terms the Kafka effect, an effect that both Hoffman and Kaplan enact through their autobiographical writing by discovering the extent to which they are, as language in general is for Deleuze and Guattari, “unique, ... a mixture, a schizophrenic mélange, a Harlequin costume in which very different functions of language and distinct centers of power are played out, blurring what can be said and what can’t” (1986:26). Whether view ing living in translation as a privilege, as Kaplan does (1993:140), or more of a cross to bear, as does Hoffman (1989:272), both appreciate and demonstrate the difficult yet productive energy of multiple roles, voices and identities. And both learn in their autobiographical encounters with the fragments of their existence that when that whole is made of distinc tive parts in the double motion which powers the collective enunciation of the Kafkaesque machine, then the French lesson of Kaplan’s memoir is the productivity of becoming lost in translation. References Bassnett, S. and A. Lefevere (eds) (1990) Translation, History and Culture,
22
Unity in Diversity
London: Pinter Publishers. Benjamin, W. (1968) ‘Franz Kafka: On the Tenth Anniversary of His Death’, trans. Harry Zohn, in Hannah Arendt (ed) Illuminations: Essays and Re flections, New York: Schocken Books, 111-141. Canetti, E. (1977) Die Gerettete Zunge. Geschichte einer Jugend, Frankfurt: Fischer Taschenbuch Verlag. Canetti, E. (1982) Die Fackel im Ohr. Lebensgeschichte 1921-1931, Frank furt: Fischer Taschenbuch Verlag. Deleuze, G. and F. Guattari (1986) Kafka: Toward a Minor Literature, Reda Bensmaia (foreword), Dana Polan (trans), Minneapolis: University of Minnesota Press. Hoffman, E. (1989) Lost in Translation: A Life in a New Language, London/ New York: Penguin Books. Hokenson, J. W. (1995) ‘Intercultural Autobiography’, a/b: Auto/Biogra phy Studies 10, 1, 92-113. Kaplan, A. (1993) French Lessons: A Memoir, Chicago: University of Chi cago Press. May, R. (1994) The Translator in the Text: On Reading Russian Literature in English, Evanston, IL: Northwestern University Press. Nussbaum, F. (1988) ‘Toward Conceptualizing Diary’, in James Olney (ed) Studies in Autobiography, New York: Oxford University Press, 128-140. Raimund, H. (1994) ‘Wie lebt man zwischen den Kulturen?’, Die Presse, October 31, 1994. Stelzig, E. (1992) ‘Is There a Canon of Autobiography?’, a/b: Auto/Biogra phy Studies 7, 1, 1-12.
Section Two Translation in National Context
Page Intentionally Left Blank
Antigone A Scots/Welsh Experience o f Mythical and Theatrical Translation IAN BROWN
Queen Margaret College, Edinburgh, Scotland CERI SHERLOCK
Cardiff, Wales This article discusses two aspects of the nature of translation and, on the basis of that discussion, considers the nature of the Scots and Welsh languages. The first aspect is the nature of the process in which myth, specifically Greek myth, may be translated into an other culture. The authors argue that the process usually called ‘adaptation ’ is actually one of translation o f a significant mythic structure from the premises of one cultural frame to another in a way analogous to the translation of text from one language to an other. The example addressed is a version of Antigone written in Scots by Brown and translated into Welsh by Sherlock, the article’s authors. The second aspect discussed is the ways in which the lan guages under consideration are capable of dealing with the material of the Antigone play. This article discusses two aspects of the nature of translation and pro ceeds, on the basis of that discussion, to consider the nature of the Scots and Welsh languages. One of the aspects discussed is the nature of the process in which myth, specifically Greek myth, may be translated into another culture. Here, we argue that the process which is usually called adaptation is actually a process of translation of a significant mythic struc ture from the premises of one cultural frame to another in a way precisely analogous to the translation of text from one language to another. The example we address is a version of Antigone written in Scots by Ian Brown and translated into Welsh by Ceri Sherlock, authors of this paper. The se cond aspect discussed is the ways in which the two languages under consideration are capable of dealing with the material of the Antigone play as written by Brown, whose adaptation - that is to say translation of the myth brings it into a Scots cultural frame. Finally arising from this analysis it raises some tentative hypotheses about the very nature of mod em Scots and Welsh as languages. This paper results from work in progress. The classic dramatic expression of the Antigone myth is that embod ied in Sophocles’ tragedy which, it is generally agreed, was written in the
26
Unity in Diversity
period 445 to 435 B.C., while an interesting argument for 438 B.C. has been made by Geoffrey Lewis (see discussion of the background to this question in Taylor 1986). Taylor and Lewis relate the writing of Antigone to Sophocles’ participation as a general in the Samian expedition of 441440 B.C. of which Taylor (1986:xxxi) notes: Douris of Samos, a sensationalist and not totally reliable ancient source, states that the Samian prisoners were tied to posts and left for ten days, before being clubbed to death; and that their bodies remained there, unburied. The hypothesis is that the writing of the Sophocles play arose from his experience of this atrocity, containing as it does the elements of imperial realpolitik in time of war and the impiety of non-burial. In any case, it may be taken as a given that, whatever the specific stimulus to the writ ing of the play, Sophocles was centrally concerned with the conflict of religious and civic duty, of responsibilities to family and polis and of per ceived duties to the individual and the state. The myth has been returned to more than once in the twentieth cen tury. Jean Anouilh, for example, wrote a version first performed in 1944 which clearly had, however discreetly, contemporary resonance in occu pied France: in his version he presents the brave, even heroic, Antigone doing her duty in resisting the oppressive power of her uncle who is per ceived as usurping the rightful authority of Oedipus’ family. While it is perhaps this Antigone which has attracted most attention of recent ver sions, George Steiner (1984) in his Antigones discusses many other interpretations and reinterpretations of the mythic text, both ancient and modem, including those of Cocteau ( première, 1922), Honegger as op era (première, 1927) and Brecht (première, 1948). Besides Sophocles’ version, Anouilh’s, however, is most germane to our discussion here since it was through his own study of them in the original Greek and French during his secondary and undergraduate education that Brown came to the Antigone myth. Steiner (1984:177) comments on the structure of the Antigone myth: The fascination of ‘Antigone’, the pressure which the myth has ex ercised on poetics and politics, are inseparable from the presence of Creon. Antigone, herself is, in fact, absent from much of Sophocles’ play. After her exit into the night, the drama is Creon’s. Pondering the dual or ‘broken-backed’ architecture of Sophocles’ dramaturgy, commentators have repeatedly suggested that ‘Antigone and Creon’ would be a more just title.
Brown & Sherlock: Mythical & Theatrical Translation
27
The central dialectic of Antigone and the Scottish experience The key dialectic of the play between these two characters was what drew Brown to attempt a Scots version of the myth. Brown himself comes from a Calvinist background, having been, until he lapsed in 1966, a member of the Church of Scotland. His loss of faith arose from attempted censorship by a senior member of the Church of one of his earlier works, a nativity play, on the grounds that it showed Mary as ‘an unmarried mother’. The attraction of the myth, therefore, lay in the opportunity it al lowed to explore dramatically the conflict between individual conscience and acceptable received formulations of belief, the right to question and the need to maintain authority, the rebellion of the young and the con formity of the old, the dangers of unrest and the need for order in society. At a deeper level, it enabled exploration of the rights of the individual to self expression as against society’s desire to control and the potential for free thought and free will against predestinate certainty. In short the Greek myth, based on a religious and civic conflict, espe cially in Sophocles’ version, perhaps inspired - in the hypothesis of Lewis and Taylor - by historic events, offered a dialectic structure which Brown was able to appropriate to embody a new set of dialectics. Further, Brown had been drawn to Anouilh’s use of the myth to explore the French expe rience of occupation and resistance, as the Frenchman demonstrated ways in which the myth could be recreated to deal with contemporary events in a modern non-Greek culture. Brown’s dialectic, in retrospect, may be seen to be derived from the typically Scottish crisis in his own spiritual life, but at the time he began, in 1969, to write his Antigone what he was aware of was that the mythic structures allowed the embodiment of a linked series of dialectics which may be seen to be more or less specific to Scots religious and civic experience at a time when many Scots were losing interest and faith in the Church and a wider general conflict be tween the young and old was focused round protest at the Vietnam War and at oppression in South Africa. (At the time he was writing the play, Brown was one of the organizers of protests in Edinburgh at the visit of the Springboks’ rugby team). Brown, of course, is very far from alone in his experience of the dialec tical potential of Scots religious and civic experience. More distinguished writers, such as Sir Walter Scott in his Waverly Novels, James Hogg in The Confessions o f a Justified Sinner and Robert Louis Stevenson in Doc tor Jekyll and Mr Hyde, have explored aspects of the Scots psyche and created original myths which embody them. In some sense, though, the Antigone myth may be seen as an ideal template for the kind of mythic expression and exploration pioneered by such writers as Scott, Hogg and Stevenson.
28
Unity in Diversity
Dialectic and language choice in theatrical performance Brown’s dramatic form involved the interlinking of ten dramatic scenes, ten poetic choric interludes and four formal laments spoken by key char acters. As he approached writing the play, Brown chose to write in both of his languages, Scots and English, using the former for the scenes and the latter for the reflective choruses and laments. This decision to use Scots for the dramatic dialogue was in part inspired by the developing interest in dramatic work in that language which was to motivate so many Scots playwrights in the Scottish theatrical renaissance of the last twentyfive years. It was, however, inspired at a profounder level by Brown’s conviction that Scots was the appropriate language to express and explore the dialectic of the cultural conflicts his text would embody. In short, the recreated myths were about Scottish experience and, as a writer, he found that the only way he could achieve the precision and intensity of dramatic expression and direct personal experience he wished to convey was in Scots. This choice was reinforced by the emphasis he places in his ver sion on the family relationships and domestic and emotional circumstances of the characters, Scots being used in his childhood experience more of ten in domestic and family contexts, English being used in more formal contexts. The fact that the laments and choruses were in the 1969 version written in English illustrates the general perception held, at least then, that that language might be more apt for the expression of more reflective and emotionally dispassionate material, a view which would support Manfred Gorlach’s disputed theory of diglossia in the use in Scotland of Scots and English. In the event, when the play was revised first in 1992, Brown put the choruses and laments into Scots, recognizing that in fact the Scots language was capable of the levels of reflection and dispassion he had been educated in the fifties and sixties to think impossible. The stimulus to the revision of his Antigone was the co-operation of Brown and Sherlock on Bacchai, a translation of Euripedes’ The Bacchae presented by the Welsh company, Dalier Sylw, at the Cardiff Interna tional Festival in October 1991. In this production, the dramatic scenes were played on alternate evenings in Welsh and English, Brown being the translator into English, while the choruses were played throughout in the original Greek. The production was underscored by music composed by Michael Rosas Cobian. The production, while exploring intertextuality in performance, had remained clearly focused in its aim of achieving a theatrical impact derived from attention to a fundamentally classical read ing of Euripedes’ work. The collaborators found the dramatic and mysterious effect achieved by the Greek choruses and the differing im pacts and effects of the two dramatic texts fascinating and sought to explore further their theatrical collaboration. Brown told Sherlock of his Scots
Brown & Sherlock: Mythical & Theatrical Translation
29
language version of Antigone and proposed that, once Brown had carried out some revision, Sherlock might consider exploring Brown’s play both in Scots and translated in Welsh, experimenting in developing it into an operatic form, and establishing its future potential through workshops and productions mainly directed by Sherlock. The authors discussed their views of the nature of their languages, pursuing the hypothesis that, with different histories and political contexts and arising from different IndoEuropean language groups, they might address the same topics, emotions and concepts in different ways, achieving and communicating their lin guistic implications differently. This project started in 1993. For Sherlock, the fascination of the trans position of a mythic worldview from a theatre text into a performance context was of prime importance. How could the authors transpose and represent the world view of fifth century Greece into a coherent, playable, and intelligible live experience? Moreover, how could they translate this playtext with its roots of performance, world view, and structure firmly embedded in a different and distant past into an actable dramatic event? For Sherlock as director, how could he create a mise en scene? This is not a new problem and the weight of performance tradition attests to the abil ity of these texts somehow to speak supra-historically. The authors sought, however, a resolution appropriate to their present societies and languages. Another dimension to their exploration was derived from the axiom that reading a playtext is as difficult as reading a musical score.The au thors would add that re-reading that text and allowing it to be represented as a performance text is not only difficult but, philosophically, an impos sible exercise. Patrice Pavis (1992:155) put this dilemma more dramatically when he notes that: the text is much more than a series of words : grafted on to it are ideological, ethnological and cultural dimensions. Culture is so omnipresent that we no longer know where to start investigating it. Despite these difficulties, however, the authors would argue that there is an imperative implicit in the text that demands performance, that the text itself requires performance to propel it into coherent intelligibility. Here, though, is the nub - and this was the enquiry of Bacchai and the basis of its dramaturgy as a performance text - the Greek myth’s playtext lan guage was overlaid and obscured by a collective memory. This meant it was no longer as clearly discernible as it originally was. This fact urged us, through its implicit textual imperative, to seek to give it birth and new form. Scots and Welsh in translation In the Bacchai project of 1991, our resolution was to create a site-specific
30
Unity in Diversity
staging, taking the audience through an experiential journey with Diony sus and the Bacchantae, giving the narrative drive in accessible direct language with the choruses, like some dimly recalled songs in an un remembered language. Brown provided, in a simple yet pithily resonant language, a translation of the scenes which offered a foundation for the stylized, articulated Greek of the choruses. Brown’s language was spare, derived from Scots usage, while the Greek choruses seemed like the most tribal Welsh, or rather Celtic, dialect. The audience understood both. In retrospect, and this is where the Antigone project began for Sherlock, the dramaturgical team were left with the question of whether the choric lan guage could also be translated and retain its impacted nature? Could a new cultural context or adaptation of this playtext provide a more intel ligible performance text of a Greek text? In Bacchai, the Welsh translation was prepared by Gareth Miles. While that translation had remarkable qualities, it highlighted a fundamental problem about translation into Welsh which Brown’s translation had been able to avoid. This problem was that of finding an appropriate modem register. So stimulating were the implications of this revelation that, at Brown’s request, Sherlock opted to act as translator of Brown’s Scots text into Welsh in order that he might understand more directly and fully the issues being examined and explored. Brown’s extant Scots language version of the Antigone myth had avoided similar problems to those the Welsh translation used in Bacchai experienced with the types of language inherent in the Greek playtext. This it did by recreating an original text. This used the structural model of the Greek myth to accommodate and mediate Scots concerns and mythological structures. His solution, however, was not the only way Scots languages could accommodate translation of Greek mythic mate rial. Of the two versions of Bacchai, Welsh and (Standard Scottish) English, it was the latter which was generally considered to be the more success ful, fluent and fluid. In his earlier draft of Antigone, because, as we have argued, of the long held traditional belief that Scots could not adequately articulate the language of philosophical debate and speculation, Brown had had difficulties in accepting the range of modern Scots; after his ex perience of the Bacchai project, he came through understanding the potential of Standard Scottish English to accept the potential of present ing choruses in Scots. Meanwhile, Sherlock found profounder difficulties in translating into Welsh. There is an implicit historicism, datedness, that dogs and grounds the traditional mode of ‘poetic’ translation in the Welsh language, tradi tionally a kind of ‘High’ Welsh invented, and it must be said perfected, by Wales’s greatest internationalist dramatist, Saunders Lewis, and also in the definitive translation of Antigone by the poet and academic W. J.
Brown & Sherlock: Mythical & Theatrical Translation
31
Gruffydd (1950). Both playwright and poet had an explicit agenda in their work of translation, returning Welsh literature to the aruchel (prince ly or chivalric) world of the Welsh early mediaeval poets, the Gogynfeirdd. For them, Greek theatre represented a literary golden age which needed to be married to a similar period in Welsh literary history and, by so do ing, give it status, integrity, literary merit and intelligibility, even if for the few. For Lewis, the agenda was even more political, to return Welsh literature, and implicitly Welsh culture, to a Catholic age, to an age of linguistic and literary sophistication, and to a mindset far more expansive than the narrow polarities of Protestantism, and the literary fundamental ism of nonconformity, which had Wales and Welsh culture in such a grip until very recently. This phase of our research was concerned with the ways in which the two languages under consideration, Scots and Welsh, are capable of dealing now with the material of the Antigone myth. It was based very much on that material as recreated by Brown, whose adaptation - that is to say translation - of the myth, as we have seen, brought it very much into a contemporary Scots cultural frame. The use of modem Scots al lowed Brown to deal with the matter in a way which allowed large scale questions of political power and personal duty to be explored and yet to present these issues within a more intimate and domestic frame. This may result from the history of Scots since 1603. While it has retained a function and capacity in dealing with important religious and political matters, given the hegemonic authority of English or Standard Scottish English, to use A. J. Aitken’s phrase, such usage has been more prevalent in domestic contexts. Given that, it may be that a fluency and ease of tran sition from register to register in Scots and between Scots and various forms of English is very much part of the Scots language experience. Such a process would explain the theory of Scots/English diglossia and, specifically, the approach Brown first took to presenting the choruses in English, not Scots, when it is evident that in fact Scots is more than capa ble of dealing with such material. The dramatic flexibility of modern Scots Some examples may exem plify the points being made about the expressivity and fluid capacity of Scots to deal at the same time with both large and intimate issues. In Scene One of the play, Etiocles and Polynices are arguing about Etiocles’ refusal to fulfil the condition of sharing au thority in the state: Etiocles: Ach christ. Aw ye want is pooer. An yer ain wey. An ye cover it up wi fine wirds and fancy ideals. Ye couldnae staun it that
32
Unity in Diversity A wis the yin that faither looit. Ye kidnae staun his bein gid te me. Ye’re sick-green wi envy an greed. Polynices: Oor ower gid brither disnae think A’m suitit te rule. He conseeders A’m ower loose in ma weys (Brown 1996).
Here the directness of the language embodies argument about power structure at the same time as about family affection and hierarchy. The language avoids polysyllabic expression and allows hidden references not available in English; for example, the term ‘gid brither’ is Scots for brother-in-law, a hidden pun within Polynices' reference to his ‘ower gid brithe’. Again, when the children’s nurse talks, after Polynices’ death, of his dissolute behaviour, she is able to do so through very precise, intimate and indeed affectionate language which nonetheless firmly suggests a clear moral frame which Polynices has transgressed: A ayewis telt Polynices he’d get hissel inte trouble. Ye cannae ayewis be lauchin A seyit te him. Ye’re a braw ladddie noo, bit ye’ll soon be gettin wee pleuchs roon places ye dinnae want te see them. This fluidity is reinforced by the easy movement available between registers in Scots, which has, of course, a large shared vocabulary with English, especially in Latinate and other classically derived terminology, some of which, indeed, like ‘incarcerate’ and ‘missive’, on the evidence of no less an authority than Samuel Johnson, according to James G. Basker (1993:83) has entered English vocabulary through Scots. This facility is seen in the following dialogue bteween Creon and Antigone from Scene Eight which leads in this version to Antigone’s defiant confession to Creon of her act of rebellious piety: Antigone: An A’m te be a figure in the reconstruction plan? Creon: Aw lass. A’m only tryin te dae ma best. Antigone: Whit fir noo? Creon: Ye speir a lot o questions fir a lassie. Antigone: Sae ye could hae yer granweans. Whit wey no some ither wife? Creon: Ye’re Oedipus’ dochter. We need peace in the end. Antigone: A fine poleetic maiirriage. Creon: An he dis loo ye. Antigone: A’m Oedipus’ dochter. An the blid is bad. The changes arising from the choice of Scots rather than English are not only in the area of cultural implication discussed here, but also in the areas of rhythmic and harmonic expression. That this is so may be made clear by the following example of the transformation effected by Brown’s
Brown & Sherlock: Mythical & Theatrical Translation
33
decision to rewrite the choruses in Scots. Chorus Seven in the 1969 ver sion went as follows: The grey light of dawn is flying away through the pink low sky To the deep clear blue with the white mountain clouds and their flattened bases It is fresh and beautiful and sharp And here we watch Afraid The Scots text is as follows: The grey licht o dawn is fleein awaw through the pink laigh lift Te the deep clear blue wi the white mountain cloods an their flattent bases It is fresh an braw an shairp An here we watch Afeart Sherlock observed that Brown’s new playtext used the Scots in a more paradigmatic way than his (Scottish) English rendering in Bacchai and that moreover Brown’s (Scottish) playtext had now become a richer performance text whose full weight could only be realized with vocal articulation. (An oral text whose playing requires an aural reading). The semantic and syntactical nature of Scots with its oral and dialectal rhythms, vocabulary, debating tradition and social hierarchy suited bet ter Brown’s goal of replacing the Greek world-view with that of the Scots. For Sherlock, this in particular explains the fluidity with which Brown describes his text as at once both a translation and an adaptation. Brown had in fact transposed the text into Scots. The possibility of such a linguis tically dynamic approach recalls the avant garde performance experiments of the late sixties, more especially the work of Richard Schechner. In talking about the Performance Group’s famous Dionysus 69, Shechner (1988:70) notes: the ‘scripts’ I am talking about are patterns of doing, not modes of thinking. Talking does not appear first as configuration (words-aswritten) but as sound (breath-noise). Ultimately, long after writing was invented, drama arose as a specialised form of scripting. What Brown had discovered in his transposition of the Antigone myth into Scots was that, more than English, Scots functioned as a gestural language and therefore one which fitted the dramaturgical imperative of the project.
34
Unity in Diversity
Welsh and its mythic and dramatic scope For Sherlock, the question was whether Welsh, with its highly develop ed literary tradition honed over thirteen centuries, functioned in the same way. Certainly, dialect in Welsh exists as much as, if not more than, in Scots: it is said that dialects in Wales can place people to the nearest farm or conurbation. They are, however, also extremely inscaped by geo graphical, economic and geopolitical implications. The result is that they are rendered more parochial than epic. Cheek by jowl with the Welsh language, particularly in the more industrial and metropolitan areas, there is also the wide use of Anglo-Welsh whose uneasy relationship with Welsh generates active tension, even given its espousal by great voices like Ed ward Thomas, Harry Webb, Tony Conran and, of course, Dylan Thomas. A native, albeit bilingual English/Welsh, speaker like Sherlock has in herent personal prejudices to surmount before considering it a language in its own right rather than the offshoot of the slippage of Welsh against the attempts at cultural suffocation of Wales by a more dominant neigh bour. Anglo-Welsh does have a distinguished history of literature and a broad and appealing communicability. Nevertheless, in Sherlock’s con sidered opinion, its attributes seem to suggest it to be a sophisticated and highly developed dialect and not a language, as Scots, in view of his experience of it, clearly is. The elision of the Germanic, Celtic and, later, Romance roots in Scots clearly illustrate a language culture at a very different level of its devel opment, and one which works in a more germane fashion with the core of our dramaturgical enquiry, namely the transposition of one playtext into another language. Welsh itself, with a more classical language structure and historico-literary tradition, offered a more limited possibility where world view and culture were concerned. Again, when one turned to the syntactical polyrhythmic accents of the Greek, Welsh with its constant history of strict metre verse, both oral and written, was another story from Scots. It was possible to preserve the ‘poetic’ nature of the Antigone text, though here the use of cynghannedd (strict metre verse poetry) forced a ‘closure’ on the text in a way that the Scots resisted. Tentative interim conclusions There are a number of hypotheses which may be adduced to explain the flexibility for our purposes we discovered in modern Scots. Firstly, it lacks a standard dialect with undisputed authority. One of the results of this is that the thirteen or so recognized regional dialects are each seen as having their own integrity, their prestige being acquired from a variety of factors, the dialectic, for example, of modern/urban versus old-fashioned/
Brown & Sherlock: Mythical & Theatrical Translation
35
rural being only one within which prestige may be established - and, given the nature of Scots economy and geography, it is easy to see that it is not self-evident which aspect of such a dialectic would have the higher pres tige. A second result of this is that the language is constantly developing, not only through neologisms like ‘sitooterie’ for ‘patio’, but by borrow ing between dialects: the introduction of a recent word like ‘scoosh’ (something easily achieved) from the Glasgow “patter” into other dialects is only one example of this process, while, equally, older words return to use as in the case of ‘bourach’ (a messy heap) more frequently used in North-eastern Doric in recent years. Secondly, Scots happily acquires new words from new cultural dimensions, such as film or pop music, in common with most other languages, developing from such specialised borrowings it own lexis: ‘segue’ in three centuries may have its own us age in Scots as the purely Scots lexical item, ‘demit office’ (formally stand down from an official position), now does. Thirdly, although its relationship with its sister language English is notoriously uneasy, to the extent that many distinguished commentators have argued that Scots is now a dialect of English and some even that it was never a fully inde pendent language, these historico-linguistic views must, however, be contrasted with the everyday experience of many English speakers who, in terms of everyday intercomprehensibility, find Scots a distinct lan guage. Nancy Banks-Smith, for example, recently (1996:26) observed, “I have discovered that no-one south of the border understands a word that Rab C Nesbitt (BBC2) says”. Indeed, it is the experience of Sherlock, as a speaker of Welsh, English, French, German, Italian, Latin, Greek and Spanish, that his approach to translating Scots is as distinct as that of working between those other languages. Nevertheless, the closeness of syntax and lexis and the interchange of words and nuanced meaning be tween Scots and English can be a strength to both. Finally, Scots culture may be said to be imbued with an interactive democracy, evidenced, in ter alia, by its systems of established church government, its historic system of royal election based in a high king principle, leading to the title ‘King/ Queen of Scots’, not ‘of Scotland’, or, even, to be a little controversial, its peaceful, non-violent and parliamentary reaction to suppression of the votes of the majority in the 1979 Devolution Referendum. This culture of constant and, often, domestic debate provides a language which is capa ble of the combination in one register of the levels of public and private, large and small scale, already discussed in this paper. It is clear that a ‘straight’ translation of the Antigone myth direct from the Ancient Greek in the terms of our dramaturgical investigation is not possible. In translation into Welsh, as has been argued, the weight of literary language and the slightness of dialect tend to impose a seemingly predetermined and restricted end result. Egil Tornquist (1991:170), an
Unity in Diversity
36
advocate of transposition in the way in which Brown’s playtext works, although, to be sure, he is using it to illustrate transposition from one me dium to another, makes an important point: there is the challenging fact that it is by examining drama in trans position, rather than by studying the text on its own, that we come to perceive what this hybrid artform is all about. The challenge therefore for Sherlock’s translation is to find an appro priate linguistic modus operandi to represent the essence of the original playtext in a language which can both assimilate the register of the ori ginal and provide the apposite dialectical structure of the performance text in Welsh. Charles Segal (1986:370) notes: The process of interpretive understanding is a shifting movement between recognizing the text in its unassimilable otherness, its ulti mate strangeness, and making the text in some sense our own, something which we can assent on the basis of our experience of what the text signifies. Future stages in the project The dramaturgical consequence of this investigation is no longer to at tempt a balance of the synchronic and diachronic tensions implicit in the text, but to move beyond them by taking what in essence the drama is based on in Antigone, namely character and civil/personal conflict and propelling them into a situation which will allow them to communicate. That is to adapt the source playtext. Brown’s inspiration was, of course, the Anouilh version. His own work has taken on this idea and explored the concept and practice of transposition as translation of myth. The un expected and fresh result is foreshadowed by Pavis’s (192:156) comments on the nature of translation: Translation is the undiscoverable mythic text that tries to take ac count of the source text - all the while knowing that such a translation exists only with reference to a source-text-to-be-translated. Added to this disturbing circularity is the fact that theatre translation is never where one expects it to be: not in words, but in gestures, and in the ‘social body’, not in the letter, but in the spirit of a culture, ineffable but omnipresent. The exploration of the question of translation of the Antigone myth into Scots and Welsh and from Scots into Welsh continues. Brown and Sherlock have scheduled workshops with Scots and Welsh speaking actors to explore further the potential and significance of their texts. They also
Brown & Sherlock: Mythical & Theatrical Translation
37
seek to explore at a later stage translating the dramatic text into operatic form. This paper has sought to present an interim report with conclusions, some a little tentative, but many the present authors consider remarkably firm, about the ways myth may be translated into another culture, the aspects of modem Scots and Welsh capable of dealing with the material of the Antigone play, and, perhaps most to our surprise, but most rewardingly, a deeper understanding of the very nature, as living lan guages, of Modern Scots and Welsh.
References Banks-Smith, N. (1996), The Guardian, January 27, 1996. Basker, J. G. (1993) ‘Scotticisms and the Problem of Cultural Identity in Eighteenth-Century Scotland’, in J. Dwyer and R. B. Sher (eds) Sociabil ity and Society in Eighteenth-Century Scotland, Edinburgh: Mercat Press. Pavis, P. (1992) Theatre at the Crossroads of Culture, trans. Leon Kruger, London: Routlege. Schechner, R. (1988) Performance Theory, London: U.P. (Revised edition). Segal, C. (1986) Interpreting Greek Tragedy, Ithaca, New York: Cornell U.P. Steiner, G. (1984) Antigones, Oxford: Clarendon Press. Taylor, D. (1986) Sophocles: the Theban Plays, London: Methuen. Tomquist, E. (1991) Transposing Drama, London: Methuen.
Page Intentionally Left Blank
‘Genuine’ and ‘Fictitious’ Translations of Science Fiction & Fantasy in Hungary ANIKÔ SOHÂR
CETRA Leuven Research Centre for Translation, Communication and Cultures, Katholieke Universiteit Leuven, Belgium This article examines the way in which science fiction and fantasy novels are developed and established in Hungary through transla tion proper as well as through a system of fictitious ’ translations, a special case of the pseudo-translation phenomenon in contempo rary Hungarian society. It outlines how Hungarian translators and publishers import books and translational solutions in order to es tablish and promote a genre while using more general patterns even to the point of creating fictitious foreign texts. In this article I shall examine the way in which science fiction (SF) and fantasy novels are produced and established in Hungary through transla tion proper as well as through a system of fictitious translations, a special case of the pseudotranslation phenomenon in contemporary Hungarian society. I would like to demonstrate how Hungarian translators and pub lishers import books and translational solutions in order to establish and promote a genre while using more general patterns even to the point of creating fictitious foreign texts. Firstly, it has to be pointed out that before the political changes in 1989/1990, Western-type popular genres hardly existed. For the commu nist value system, literature for the purposes of entertainment was poison to the mind. Nevertheless, ‘light reading’ was published even during that period, especially detective and spy stories. From 1969 on, a publishing house specializing in children’s literature (Mora Ferenc Konyvkiado) had a science fiction magazine as well as a science fiction series, publishing from four up to twelve books annually. Other publishers also had one or two novels of ‘fantastic literature’ translated, but the number of science fiction publications was insignificant. This situation has changed radi cally. Several publishing houses specializing in popular genres have come into being since 1989 and the number of SF novels published each year exceeds eighty.1 The contribution of Hungarian science fiction writers (with or without pseudonyms) was relatively high during the first two years, but it has been decreasing since then.2 According to the Hungarian
1 In 1989 (80); 1990 (107); 1991 (88); 1992 (94); 1993 (107); 1994 (89). 2 1989:30 novels (11 pseudonyms); 1990:44 (17); 1991:19 (6); 1992:13 (7).
40
Unity in Diversity
National Bibliography and my own investigations, 1995 appears to be the first year when all the published SF novels written by Hungarian au thors were either pseudo- or fictitious translations.3 “It is texts which have been presented as translations with no corre sponding source texts in other languages ever having existed - hence no factual ‘transfer operations’ and translation relationships - that go under the name of pseudotranslations, or fictitious translations”. This defini tion, formulated by Gideon Toury (1995:40), describes the phenomenon adequately; nevertheless, it seems serviceable to distinguish pseudo translation from fictitious translation. The term ‘fictitious translation’ will be used in cases when the author acts as the translator of an English ‘original’ and fictitious bibliographical data are also provided: for exam ple English publisher, English title rarely corresponding to the Hungarian one; date of publication; date of translation. In some cases a preface or an epilogue by the novelist confirms the illusory foreign origin. This method was introduced in a fantasy series, M.A.G.U.S. or the chronicles of ad venturers, launched in 1992 by one of the first publishing houses to specialize in science fiction and fantasy, Valhalla Paholy. Since then it has become fairly widespread. These fictitious translations work so well that some novels (e.g. those allegedly written by ‘Wayne Chapman’ and ‘Martin Clark Ashton’) have been listed among the translated English/ American books in the Hungarian National Bibliography. The term ‘pseudotranslation’ will be applied when only a pen name, and possibly a fictitious English title, are provided. “The interesting paradox is that this classification can only be applied after the veil has been lifted. Consequently, texts can be approached and studied - as pseudotranslations only when the position they were intended to have, and once had in the culture which hosts them has al ready changed” (Toury 1995:40). However, this is not the case in Hungary, most of the pseudotranslations and all the fictitious ones are considered to be genuine translations, and it is therefore possible to study these texts while they are still functioning according to their author’s intention. The phenomenon of fictitious translation in Hungary is obviously re lated to the relative novelty of science fiction and fantasy. Toury (1995:42) points out that “the decision to present a text as a translation, let alone compose it with that aim in mind, always suggests an implied act of sub ordination, namely, to a culture and language which are considered prestigious, important, or dominant in any other way”.
3 The use of pseudonyms seems to be increasingly systematic in all popular gen res. I have already found 372, mostly English pen names, between 1989 and 1995 (45 in SF and fantasy). It seems that a (preferably English-sounding) pseudonym is demanded by the publishers.
Sohar: Translation of SF in Hungary
41
It is an interesting fact that the first attempt (1989) to introduce fan tasy to the Hungarian reader was rather unsuccessful; the publishing house in question (Unikornis Kiado) gave up the enterprise4 after the twelfth published novel and turned to classic, canonical literature. Never theless, they were the first to publish fictitious translations of ‘John Caldwell’ and ‘Wayne Chapman’, who happen to be the most popular fantasy writers in Hungary at the moment. It is worth mentioning that the first Hungarian fantasy (Chapman) and the first parody of the genre (Caldwell) emerged at the same time. Before analyzing these fictitious translations and indicating the com ponents underlying their success, it seems necessary to summarize the general translation policy of the publishers who specialize in popular fic tion. I will concentrate on the two most important groups5whose approach appears to be most characteristic. They claim publication of one book per week, mostly in science fiction and fantasy. This is approximately one third (230 volumes) of the entire SF and fantasy production in Hungary for the period covered in this article (1989-1994). First of all, the principles of selection have been changed. During the communist era only politically approved novels were allowed to be trans lated, in effect mainly writers from ‘socialist countries’ or progressive Western novelists whose social concern was explicit. In the first wave of enthusiasm after the political changes, translators and publishers wished to fill a long-felt gap between the West and Hungary. Thus the most fa mous novels, such as Philip K. Dick’s Do Androids Dream o f Electronic Sheep?, Alfred Bester’s The Stars My Destination, Brian W. Aldiss’ Hot house, Arthur C. Clarke’s 2061. Odyssey Three, as well as innumerable Isaac Asimov books were translated. In a free market situation the main criterion determining the production of a book is whether it is market able. This has led to a remarkable restriction in the number of source cultures and languages. According to the Hungarian National Bibliogra phy and on the basis of my own research, only English and American science fiction has been translated into Hungarian since 1993. Another new element in the translation of SF is the generally low quality of these translations. Before 1989 the publishing houses could afford to invest the money, time and energy needed to achieve high qual ity, because they were subsidized. Today, due to the peripheral literary situation of all popular genres, their total lack of prestige as well as the 4 The series advertisement promised: IZGALOM! KALAND! FANTAZIA! SZERELEM! KOLTESZET! (EXCITEMENT! ADVENTURE! FANTASY! LOVE! POETRY!) - on the back cover of each novel. 5Valhalla Paholy (Pendragon, Neotek) is in the capital, Cherubion (Phoenix, Lapics) in the country; both of them publish popular genres exclusively, mostly fantasy, and many pseudo- and fictitious translations.
42
Unity in Diversity
publishers’ desire to maximize profit in as short a time as possible, the translators have to produce translations in a very short time, which of course results in errors (for example ‘travail’ is not translated in the pref ace of Peter Morwood’s The Horse Lord but used as a proper name),6 mirror translations (including the imitation of grammatical structure at sentence level), omissions, and so on. For instance, a translator with a science fiction profile, Zsolt Szantai, translated ten (and, using several pseudonyms, wrote four) novels in 1995, each approximately 300 pages long. In order to save money, most publishers do not employ editors or proof-readers at all, and published books are therefore full of misprints, incorrect hyphenation, unnecessarily different characters, blanks, and so on. Only those experienced translators who quickly learnt how to use a word-processor can present easily readable texts. In other words, there is virtually no proof-reading or copy-editing and the translator gives a dis kette directly to the publisher or printer. Since science fiction has a clear identity with little variety in the vari ous cultures, genuine translations will presumably be source-oriented, that is, they will follow the norms, patterns and models of the original culture, especially as there are no well-established norms of translating popular genres into Hungarian. However, it would appear that rather strict rules are observed in the case of certain publishing houses. In examining the translations, it is striking how obviously the publishers prefer a cer tain text length or format.7 These formats are fairly close to the standard English paperback SF editions. Hardbacks are the exception, though un like the English or American equivalents, the back cover usually specifies the genre (‘a new legend of the science fiction genre’, ‘already a classi cal fantasy’), that is they are labelled ‘fantasy’ or ‘SF’. Substitution of the original cover and layout is commonplace. In my research to date I have found only twelve translated science fiction books with the original cover. Edward James (1994:2) has the following to say on SF and fantasy book covers: “a fantasy novel will display dragons, elves, wizards, and heroes (or heroines) with swords, mighty thews, and not very many clothes; an SF book can draw on a futuristic machinery or architecture, robots, weird alien landscapes and alien creatures ... But most of the covers represent the core of what each genre is ostensibly about, and may suggest to you some possible divergences between gen 6 “May be by HEAVEN’S grace this land may be delivered from sore Travail” (Peter Morwood: The Horse Lord, p. 7) is rendered as “Talán a Menny kegyelmét még elnyerheti e fold Tarvail által” (‘Maybe this land may penetrate Heaven’s grace by Tarvail’; Hungarian edition, p. 1). 7For instance, the Debrecen group seems to favour relatively short novels (18 cm, 180-190 pages, 12 A/5 sheets), either genuine or pseudotranslations, while Valhalla prefers longer texts (19 cm, 300-310 pages, 19 A/5 sheets).
Sohar: Translation of SF in Hungary
43
res”. It is possible that the change from the original covers represents the way in which these relatively new categories are visualized in Hungary, at least in the mind of the publisher. The layout is also adapted to Hungarian conventions. For instance, chapters usually begin on a new page, so the publisher of the first two volumes of The Hitch Hiker's Guide to the Galaxy has chosen this more conventional page-setting, while a more source-oriented layout - con tinuous chapters - is presented in the third volume which was brought out by another publishing house. The frequency of adapted titles,8 slightly or completely different from the original, is also rather high: 153 cases out of 388. A tendency to short en the original title (40 out of 153) and to highlight an essential feature (e.g. a proper name) can be noted, for instance Robert Anson Heinlein’s Stranger in a Strange Land is rendered as ‘Annunciation’ (Angyali Udvdzlet), Stephen King’s The Eyes o f the Dragon as ‘The Eye’ (A szem), Terence Hambury White’s The Once and Future King as ‘Hail Arthur, Great King!’ (Udv neked Artur, nagy kiraly/), or Vonda M. McIntyre’s The Entropy Effect as ‘The Enterprise Takes Off’ (Az Enterprise elindul). It is interesting that a similar trend can be observed in examining the English titles of fictitious translations, for instance when a Hungarian title is Two Moons, the hypothetical English title is The Quest of Two Moons, or Worluk's Curse appears to be Hero o f Dimensions ‘originally’. Only 16 out of 46 fictitious English titles correspond to their Hungarian ‘translations’. The text itself is usually re-organized on a micro level, that is the paragraphs (in Hungarian they tend to be shorter) and the sentences (in Hungarian they are longer) often do not correspond. It also occurs rather frequently that the volumes of a trilogy or series are translated by differ ent translators or even published by different publishers without any reference to the already existing translations, e.g. identical parts and iden tical proper names are rendered differently.9 Naturally, translating implies omissions, additions, and alterations, especially in the case of Hungarian since iti structure is rather different from English and the Indo-European languages in general. Nevertheless, in many cases the actual modifications, such as the treatment of proper names, can merely be explained by the requirements of the publisher who, it goes without saying, is very dependent on the perception and recep tion of science fiction. As previously mentioned, popular genres have a 8 For instance, the Hungarian title of Philip K. Dick’s Do Androids Dream of Electronic Sheep? would be ‘Winged man-huntef in English. 9For instance, Douglas Adams’ The Hitch Hiker*s Guide to the Galaxy or William Gibson’s Neuromancer trilogy.
44
Unity in Diversity
peripheral position in the literary system; therefore, at least in principle, they are mere products for consumption, with no lasting value at all. However, the value shifts in some translated novels10indicate the oppo site, that is, some topics acceptable to English or American readers are considered controversial by the Hungarian translator and are consequent ly replaced with more acceptable subjects. Parallel to this, another trend can be observed which goes against conventions.11 This reflects the am biguous nature of peripheral genres as well as the absence of overall normative regulations. It is noteworthy that those publishers who were involved in translating before 1989 are more inclined to follow conven tional patterns, particularly when writing their own pseudotranslations, than those with no previous experience in the field. These observations hold true in the case of pseudo- and fictitious trans lations as well. It has to be emphasized again that these texts are received as genuine translations by the Hungarian reader although sometimes their language use is perceptibly more playful, more inventive, using the po tential of the language to the full. All novels of this category follow the international patterns for the genre; for instance a fantasy novel always takes place in a pre-industrial but highly magical society and inevitably comes to a happy ending, science fiction describes a scientific break through and its consequences, and so on. Cyberpunk seems to be the sole exception; so far I have not found one Hungarian cyberpunk novel which refers to cyberspace, or virtual reality, but this phenomenon needs fur ther investigation. In line with international trends, approximately 70 per cent of pub lished novels in Hungary are fantasy. Another external element, previously unknown in Hungarian writing, has been borrowed, namely shared imaginary worlds and/or borrowed characters.12 The most popular and most significant example is the previ ously mentioned ‘M.A.G.U.S.’ series which presents a highly elaborate world with sophisticated history, geography, religions and languages, 10Examples include ‘I had wives’ (translated as ‘I have a wife’, p. 192) and ‘She was the only one with the strength of character to murder me’ (translated as ‘She was the only one with the strength of character to get involved in a dispute with me’, p. 153) in Terry Pratchett’s The Colour of Magic. 11 For instance, ‘Everything You Never Wanted To Know About Sex But Have Been Forced To Find Out’ (p. 174) is rendered as ‘Everything you wanted to forget about sex but your wife wouldn’t let you’ (p. 31), and ‘The Big Bang Theory - A Personal View by Eccentrica Gallumbits’ (p. 174) is rendered as ‘The Virgin and the Twins, or menage a trois or Gruppensex in space by Eccentrica Gallumbit’ (p. 31), both in Douglas Adams’ The Restaurant at the End of the Universe. 12 For instance, in his Chaos series ‘John Caldwell’ utilizes a magician, Moon light Thief, who was invented by ‘Robert Knight’.
Sohdr: Translation of SF in Hungary
45
written by several authors, and which also serves as a basis for a fantasy role-playing game. Although fantasy as such has become astonishingly popular, the over whelming success of the main writer of fictitious translations, ‘Wayne Chapman’, has overshadowed the other writers in the genre. Zealous read ers were willing to pay a sum equal to the average monthly wage in Hungary for the first editions of his novels on the black market before their reissue. Second editions of popular literature are fairly unusual in Hungary at the moment;13 when they appear, they indicate a very po sitive reception. Two of the components underlying the success of fictitious trans lations, and particularly of ‘W ayne Chapman’, have already been mentioned: the detailed elaboration of the imaginary worlds, similar to Tolkien’s The Lord o f the Rings, and the playful use of language. An other factor could be the palpably cynical ideology manifest in these novels, although it is never explicit, concealed behind religious, mystical and/or occult occurrences. Thus the characters of the main heroes (and a heroine in one case) are always fully developed. They are never pre sented as good or evil but as ambivalent and frail (even when they have supernatural abilities), though they always fight for the ‘good’ side, and, of course, win eventually. However, this does not mean the usual happy ending since the message of these works to the reader is that life is noth ing but a continuous struggle. Only temporary victories can be achieved; neither values, nor principles endure and therefore do not merit our es teem. In fact, there is no essential difference between right and wrong. This relativistic attitude is reflected in the plots as well; for instance the well-paid saviour of the fair maiden takes advantage of the situation,14 or the servant of utter evil, a merciless assassin who cherishes human sacri fices, turns out to be the most positive character.15 This withdrawal from any ethical commitment seems related to the consequences of the recent socio-political changes, namely the general and deepening uncertainty regarding values. Nevertheless, the presence and increasing influence of mystical and occult topics, both in society and science fiction, should be perceived not only as a result of the loss of a stable set of values but also as part of the current process of cultural importation. In sum, it seems that instead of the hypothesized two approaches 13 Apart from classics, such as novels by Jules Verne, there were 19 second edi tions, 7 third editions, and one fourth edition (James Hilton’s Lost Horizon) during this period. 14See ‘Wayne Chapman’: A Halal havaban/In the Month of Death (fictitious Eng lish title: ‘Blood season’). 15 See ‘John Caldwell’: A Kaosz papja/The Priest of Chaos (fictitious English title: ‘Skandar Graun’).
46
Unity in Diversity
pseudo- and fictitious translations - there is only one model that is rel evant in both cases. Although Toury (1995:43) argues that “the decision to put forward a text as if it were a translation is always an individual one”, the Hungarian situation, which shows an overriding tendency to wards pseudotranslating in all popular genres, apparently contradicts his statement. The actual transfer involves genuine translations adapted to the ex pectations of the reader, as well as international patterns in pseudo- and fictitious translations. In the former case the translational strategies are borrowed from other genres that are well established in Hungarian; they therefore appear to be target-oriented for the most part. Since the pseudoand fictitious translations aim at an identical audience while maintaining their fake identity as translations, it is no mere accident that they use the same model, especially considering that the roles of translator, author and publisher overlap in many instances. It seems that this briefly described model in science fiction and fantasy might also be usable and operative in the case of all popular genres. If so, this could provide the basis for a general model for research on the trans lation of popular literature for mass markets.
References Adams, D. (1992a) The Hitch Hiker's Guide to the Galaxy. A Trilogy in Four Parts, London. Pan Books. Adams, D. (1992b) Vendeglo a vilag vegen, Budapest: Mora Ferenc Konyvkiado. James, E. (1994) Science Fiction in the 20th Century, Oxford & New York: Oxford University Press. Morwood, P. (1990) The Horse Lord, London: Arrow Books. Morwood, P. (1995) Lo ur, Budapest: Amon Konyvkiado. Toury, G. (1995) Descriptive Translation Studies and beyond, Amsterdam & Philadelphia: John Benjamins.
Theory and Practice Translation in India
PAUL ST-PIERRE Département de linguistique et de traduction Université de Montréal, Québec, Canada This article points to the need to contextualize translation and ex amines translation practices in India. Translation principles and methods are a function of the context in which they are elaborated and put into practice. In a plurilingual society such as modern In dia, relations between languages and between communities are realized and transformed through translation. The close examina tion of such relations makes it possible to elucidate the locations of power within and between cultures in a concrete manner.
Introduction1 A good place to begin this discussion of translation in India is with the following self-presentation by a practising translator, poet and critic, Prabhakan Machwe, which appeared in The Problems o f Translation. Machwe (in Gopinathan and Kandaswamy 1993: 305) writes: On the basis of my experience as a translator from Marathi, Gujarati, Bengali, Punjabi, English into Hindi, and as a translator from Hindi into Marathi and English for the last 45 years, I am recording a few facts. My first translation was in 1934 for the Indian National Con gress, Bombay; the translation of the Congress session President Dr. Rajendra Prasad’s speech from English into Hindi. The latest I am engaged in is the translation of my own monographs. Without wishing to generalize from a particular case, it is neverthe less interesting to draw attention to certain aspects of this self-presentation insofar as they can illustrate the context in which translation is carried out in India, a context which affects both expectations with respect to the functions of translation and the way in which methods of translation are defined. 1 This is a slightly modified version of a paper read at the national seminar on ‘Translation and Multilingualism in Post-Colonial Contexts: Indian and Cana dian Experiences’ held at Jawaharlal Nehru University, New Delhi, in January 1996. The research for this paper was carried out with the support of the Social Sciences and Humanities Research Council of Canada.
48
Unity in Diversity
The mother tongue The first point to be made concerns the source and target languages in volved: six different languages serve as source languages for this translator, and, among these, three - Hindi, English and Marathi - are also target languages. The plurilingual nature of this translator’s practice leads us to question the generality of certain assumptions about translation. One ex ample would be the principle which states that ‘one should translate only into one’s mother tongue’. Whereas this may well be perfectly accept able and even desirable in a monolingual or even a bilingual society, on cultural as well as on linguistic grounds, in a plurilingual context, such as that of India, it is at once impracticable and harmful, since it would have the effect of reducing the amount of translation actually carried out. This is so because if the principle were to be retained, the eighteen Indian lan guages recognized in the Constitution of India, along with English, would, given the necessity of translating only in one direction (into one’s mother tongue), give rise to 342 possible language combinations. The retention of such a priniciple would therefore not only unduly multiply the num ber of translators required but also impossibly complicate the training of such translators. A further difficulty with such a principle in a plurilin gual context such as that of India is that of providing a clear definition for the notion of ‘mother tongue’: is it the first language acquired and still spoken, the language of education, or the language of everyday transac tions? For a given individual, these might very well involve different languages. This is reflected in the choice of target languages in the case of this particular translator; that is, Hindi (in 1934 a language corre sponding to a geographical area and to a political choice), English (the language of colonization, and presently the principle language of higher education; a language of power, but not necessarily the only such lan guage), and Marathi (one of the languages officially recognized in the Constitution of India and, presumably, the translator’s first language which is still spoken). The three languages have different functions and roles to play, and each can stake a claim to an area of dominance. This is recog nized in the three-language formula recommended for Indian schools, the purpose of which is to distinguish between three tiers of communica tion: regional, national, and international (Mukherjee 1994:37).
The original text A second principle called into question by the practice of translation on the Indian subcontinent is that which states that translations should be carried out directly from original texts rather than from already existing translations into other languages. Once again, while the desirability of
St-Pierre: Theory and Practice - Translation in India
49
translating directly from the original can perhaps be acceded to, there is a need to contextualize such a principle and to chart its effects if rigourously applied. For in India, in fact, a great deal of translation involves the use of a filter language, most normally Hindi or English, but at times other Indian languages as well. Machwe remarks that even much of the trans lation into Hindi from non-Indian languages makes use of English as a filter language (in Gopinathan and Kandaswamy 1993:309-310): In Hindi there are very few people who have good equipment of foreign languages to translate from them directly into Hindi. How many of them have done direct translation from French, German, Italian, Spanish, Russian, Chinese, Japanese or any other foreign language? So mostly translations are from English, or through English. Such indirect translations involving a third language are also not un common between different Indian languages and constitute a practical solution to the impossibility of finding competent translators for certain language combinations. Thus, while there might be no translator with the capability of translating directly from Konkani into Oriya, translators with the ability to work from Konkani into Hindi or English, and others from Hindi or English into Oriya, are much more likely to exist. Such use of a filter language gives de facto recognition to certain languages rather than others, and indeed in this respect English continues to have the upper hand. Even the ‘Sahitya Akademi’ whose function is in fact “to work actively for the development of Indian letters and to set high literary stand ards, to foster and coordinate literary activities in all the Indian languages and to promote through them all the cultural unity of the country,” has published its own reference volumes in English.
Translation in Orissa: a survey A survey2 carried out in February and March 1995 supports the claim being made here that translation principles, or at the very least the actual practice of translation - and is it possible to separate the two? - are a function of the context in which they are elaborated or put into practice. A questionnaire was sent to 245 translators using Oriya, as either a source or target language; 53 translators responded. Of these, eleven said they worked with four or more source languages (one translator worked with eight) and only fourteen claimed one language alone as source language. As far as target languages are concerned, twenty-nine translators worked into one language, fifteen into two, five into three, and two into four (two 2The survey was done with the help of my research assistant, Santosh Kumar Padhy.
50
Unity in Diversity
translators in the survey did not specify their target languages). It should be noted that those translators working with only one target language did not necessarily work into the language which they identified as their mo ther tongue; in fact, more than half (sixteen of the twenty-nine translators) did not. Ten languages in all are represented as source languages, but amongst these four in particular stand out: Table 1: Number of translators by language3 Source Language 2 Assamese 31 Bengali 34 English 1 Gujarati 28 Hindi 1 Manipuri Marathi 3 30 Oriya 5 Sanskrit 0 Tamil 1 Urdu
Target 1 6 24 0 10 0 1 37 2 1 0
Four languages have approximately the same number of translators: Bengali (31 translators), English (34), Hindi (28) and Oriya (30). In the survey eight languages are mentioned as target languages, with Oriya (37 translators) and English (24) by far the most popular, followed by Hindi (10) and Bengali (6). These figures are interesting for several reasons, although it should be emphasized that the results cannot be extrapolated to other areas of India and that a similar survey for other Indian lan guages would have to be done. First, they underscore the lack of translation done between many of the official Indian languages; indeed, of the trans lators who responded to the survey none worked with eight of the eighteen officially-recognized languages, whether as source or as target language, and four other languages were only represented once. Given such results the necessity of using filter languages in translation becomes evident, for otherwise texts of certain linguistic communities would remain unavail able to members of others. Second, English remains considerably more popular than Hindi, both as a source language (34 translators for English as compared to 28 for Hindi) and in particular as a target language (24 translators for English, 10 for Hindi). The importance accorded English is reflected with even greater force in statistics relating to the publication 3Note: No translators listed for Kannada, Kashmiri, Konkani, Malayalam, Nepali, Punjabi, Sindhi, or Telugu as source or target languages.
St-Pierre: Theory and Practice - Translation in India
51
of translations. Out of the total of 1047 works translated into Oriya which have been inventoried,4 some 31% have English source texts whereas fewer than 11% have Hindi originals. Even if these figures do not cover the entire range of translations into Oriya, the importance of English is clear, and it is only over the past ten years that translations from Hindi have gained ground, roughly equalling those from English. Both English and Hindi, on the other hand, have fallen behind Bengali in terms of trans lations of works into Oriya: Table 2: Source languages o f works translated into Oriya Bengali Period English Hindi 1990-95 28 15 16 1980-89 110 85 29 1970-79 51 101 25 21 1960-69 50 25 Since 1980 more translations have been done from Bengali into Oriya than from English and Hindi combined. This leads us to our third obser vation. The importance given Bengali in the survey of translators, in particular as a source language (31 translators), is worth noting but hardly surprising given the geographical proximity of Orissa to Bengal as well as the cultural, political and historical ties between the two regions. But more than the number of translators working with Bengali as a source language, what should be emphasized is the disparity between this number (31) and that of translators working in the opposite direction, from Oriya into Bengali (only 6 translators). The inequality in the relations between these two groups which such figures suggest is borne out by the number of works translated between the two languages. Whereas, according to our figures, a mere twenty-one works have been translated from Oriya into Bengali, some 210 have been translated in the other direction.
Translation in India: the National Book Trust While it is difficult to obtain statistics concerning works published for India as a whole, it is of interest at this point to examine publications sponsored by the National Book Trust (NBT), an institution founded in 1957 and whose mandate is “to produce and encourage the production of good literature, and to make such literature available at moderate prices”. The National Book Trust publishes different collections, of which the 4The inventory is at this stage incomplete, and the tendencies would have to be verified.
Unity in Diversity
52
three most popular are: the Nehru Bal Pustalalaya series whose aim is to create “a treasure-house of enjoyable and informative literature which children can read at their own initiative” and to promote “national integ ration by providing common reading material in their mother tongue to children all over India” (more than 2700 titles); the National Biography series, which “recounts the lives of eminent Indians” (110 biographies with their language translations, totalling 700 titles); and the Aadan Praden (‘exchange’) programme, “a series of special importance because of its unique potential for forging national integration through the exchange of creative literature” - the series, “presents well known literary works, including novels, plays and short stories, of one Indian language to the people of other linguistic regions” (800 titles in twelve languages). It is clear from such statements that the role of the National Book Trust with in India is to foster a sense of national identity, precisely by facilitating communication between the different linguistic communities which make up the country. As might be expected, translation is given a major role to play, and indeed, for the six-year period from 1988 to 1994 more than 76% of the new publications by the National Book Trust consisted of translations. But despite the role of the NBT in promoting national unity and its use of translation to counter linguistic and cultural difference within the country, certain languages are given more importance than others, an importance however which does not necessarily correspond to the number of speakers of the languages in question: Table 3: Publications by the National Book Trust5 Language Assamese Bengali English Gujarati Hindi Kannada Malayalam Marathi Oriya Punjabi Tamil Telugu Urdu
% of population 1.50 8.32 .30 5.36 42.68 4.16 4.17 8.02 3.73 3.18 7.20 8.21 5.66
no. of Books 162 253 461 169 852 21 134 482 155 244 113 207 92
% of Books 4.6 7.2 13 4.8 24 6 3.8 13.6 4.4 6.9 3.2 5.9 2.6
5 Population figures for the most part are those of the 1981 census; certain figures however are from the 1971 census.
St-Pierre: Theory and Practice - Translation in India
53
Thus whereas the figures in the 1981 census for native speakers of Marathi (8.02% of the population) would lead us to expect some 285 new and reprinted volumes by the National Book Trust in this language be tween 1988 and 1994, there were in fact 482, a figure corresponding to 13.6% of all the works. Conversely, the southern Indian languages of Tamil and Telugu, with 15.4% of the total population, are underrepre-sented with 299 volumes (8.4% of the titles) for the two languages. But the pub lication figures do not simply represent what one might suspect to be an underlying north-south bias since another southern Indian lan-guage, Kannada, is overrepresented in the sample by almost 50%; although Kannada has half the number of native speakers of Telugu, more volumes have been published in Kannada by the National BookTrust than in Telugu. Thus the significance of these figures requires in-terpretation, taking into account local contexts - availability of translators, for example, and cul tural traditions - as well as historical relations between languages and communities in India. Such relations and contexts continue to exist in modern India and to influence cultural production, of which transla tions are a notable part. They are as much a result of colonial policy - the formation of a unitary state out of a plurality of princedoms, feudatory states, etc. - as of the decision to maintain the divisions in modern India along linguistic lines. Thus India is not only a state in which linguistic divisions are maintained, but it is also a nation in which such divisions can lead to new rivalries or perpetuate old ones.
Local contexts: Bengal and Orissa We have already noted the preponderance of translations from Bengali into Oriya and the scarcity of translation in the other direction. This re flects the hegemony of Bengal within Orissa, already remarked upon by one of the greatest of Oriya writers, Fakir Mohan Senapati. In his autobi ography, My Times and /, translated by John Boulton, Fakir Mohan remarks at various times on the conflicts between Oriyas and Bengalis. He writes about one such instance around 1860 (1985:40): Bengalis had all the higher, non-European posts in every single Government department in Orissa. Their sole aim was to abolish Oriya and introduce Bengali in their departments. Whenever a post fell vacant, they strove to get a Bengali appointed. And the intensity with which such rivalry and hegemony was felt by Oriyas is made clear in Fakir Mohan’s recounting of a meeting with a Bengali preacher (ibid:42): Some time later a Brahma preacher named Isan Candra Basu came
Unity in Diversity
54
to Balasore from Calcutta. I made his acquaintance and we began daily discussions on religion. After a while he switched from reli gion to language and tried to prove the excellence of Bengali and the inferiority of Oriya. This greatly angered me, so I stopped see ing him. Such relations between linguistic communities do of course affect the role given to translation. Pranati Pattanaik, in a recent doctoral thesis on The Art of Translation’ (1993), emphasizes the political nature of this role within the Indian context. The role can be that of reducing the gap between the divine and the profane through the translation from Sanskrit to the colloquial languages and thereby freeing the scriptures from the monopoly of a particular caste (1993:29-30), that of limiting British in fluence during the colonial period by the translation of other European or Indian literatures (ibid:31), or - “in the postcolonial period” - that of translating one’s own work into English in an attempt to gain increased status, English being the language in which status is obtained (ibid).
Translation theory in context After having looked at some of the effects of the plurilingual nature of Indian society on the way in which translation is carried out, it is time to turn to the theorization of the translation process. I will briefly examine here the notion of transcreation, put forward by P. Lai and which is charac terized by Sujit Mukherjee as a “method of seeking maximum readability within the confines of faithful rendering” (1994:6). Situated somewhat closer to adaptation or imitation than to translation proper (however the limits of such a term might be described), transcreation is a method which is frequently referred to in the Indian context, not only in relation to mod em versions of ancient texts and despite - or is it because of - the vagueness inherent in the notion, which has not yet received a clear definition. Thus in Transcreation: Two Essays, published in 1972, Lai does not define the notion but rather situates it within a context in which the emphasis is squarely placed on the target audience, and more particularly on the trans lator (1972:1): I soon realized that an excessive absorption in the milieu and tradi tion of English was divorcing me from the values that I found all around me as an experiencing Indian, so I undertook the translation of Indian - in practice, mostly Hindu - sacred texts, in the hope that the intimacy that only translation can give would enable me to know better what the Indian ‘myth’ was, how it invigorated Indian litera ture, and what values one could pick up from it that would be of use to me as an ‘Indian’ human being and as an Indian using a so-called
St-Pierre: Theory and Practice Translation in India -
55
foreign language, English, for the purposes of writing poetry. This passage is particularly telling insofar as translation - transcreation into English arises out of a desire to ground oneself more fully in the source Indian culture. Such an approach, as paradoxical as it may seem, is not at all foreign to the Indian tradition, which, as Pranati Pattanaik points out, has considered translation mostly as ‘new writing’ (1993:116), to the point where translations are considered as original works. Pattanaik cites the examples of translations of the Mahabharata and the Ramayana into Kannada and Tamil, and Sujit Mukherjee gives the following ac count to illustrate the lack of distinction between translation and original creation (1994:83): The Oriya poet Sitakanta Mahapatra once explained an image in one of his poems by telling me about an episode in the Mahabharata. [...] I could not recall this episode, but the explanation immediately heightened the image for me in Sitakanta’s poem. We have obvi ously been brought up on different Mahabharata texts, he on the Sarala Dasa version in Oriya, I on the Kasiram Das version in Bangla. What is of interest here is not only the way in which translation recre ates the original texts, but also that such transcreations in a sense have their own lives and lead to different cultural and literary traditions. Trans lation serves in such cases to lead to the development of these traditions, retaining and reinforcing their specificity, while at the same time main taining a certain degree of commonality. Such a compromise - autonomy from the original, and yet also reproduction - is perhaps a necessity in a country in which eighteen languages are given official status, in which no pan-Indian language has as yet evolved, in which English retains pride of place despite the ambiguities which surround its use.
Conclusion Despite the necessarily partial nature of this article it has nonetheless been possible to point to the need to contextualize the practice of transla tion, not only as it is carried out in India, but also insofar as all values and expectations as regards translation are concerned. Translation in the postcolonial context of India also underscores the connection of transla tion to power: relations between languages and between communities are actualized and transformed through translation; translation strategies reproduce more than mere meaning. The close examination of such rela tions and strategies makes it possible to elucidate the locations of power within and between cultures in a concrete fashion, and this should, it seems to me, be one of the goals of translation studies.
56
Unity in Diversity
References Gopinathan, G. and S. Kandaswamy (1993) The Problems of Translation, Allahabad: Lokbharti Prakashan. Lai, P. (1972) Transcreation: Two Essays, Calcutta: A Writer’s Workshop Publication. Mukherjee, S. (1994) Translation as Discovery, Hyderabad: Orient Longman, second edition. Pattanaik, P. (1993) The Art of Translation, Unpublished Ph.D. thesis, De partment of English, Utkal University. Senapati, Fakir Mohan (1985) My Times and I, trans. John Boulton, Bhubaneswar: Orissa Sahitya Akademi.
Marginal Forms of Translation in Japan Variations from the Norm
JUDY WAKABAYASHI Department of Asian Languages and Studies, University of Queensland, Australia An examination of the boundaries between what is conventionally regarded as translation and peripheral forms of translation such as adaptations, imitations and pseudo-translations may help clarify the nature of ‘p rototypical' translation. This article focuses on two non-prototypical methods of rendering foreign texts into a form comprehensible to Japanese readers and examines the degree of acceptance in Japan of these practices as translation. The first method, kambun kundoku, uses grammatical indicators and marks to indicate word order so as to allow Japanese readers direct ac cess to Chinese texts. The second method, common throughout Japanese literary history, involves the production of adaptations, whereby both traditional Chinese tales and European works are adapted to varying degrees, frequently by famous writers who have used adaptations as a stimulus for their own creative activities.
Introduction This article will examine two marginal forms of translation in Japan, a country whose translational traditions evolved for many centuries in com plete isolation from translation practices in the West.1 The underlying conceptual framework adopted here is that proposed by Toury (1980:43), who defines a translation as anything which is accepted and functions as a translation in a particular recipient community, and who suggests that different communities may have differing translational norms and that there is a variety of potential relationships between a text and its assumed source. The two practices discussed here may help elucidate the concept of translation in Japan, where the boundaries between translation, adap tation, imitation and even plagiarism are often quite blurred. Both the practice and products of translation have held an important position within the Japanese literary tradition, being both highly preva lent and accorded considerable prestige as the vehicle of new ideas, literary forms and expression. Even-Zohar (1978:22) has argued that the more 1Because of space restrictions, I will focus on only two marginal forms of trans lation -kambun kundoku and adaptations - although other forms such as abridged versions, pseudo-translations and diachronic translations also merit further study.
58
Unity in Diversity
important the position that translations occupy in a community’s litera ture, the more blurred the lines between translations and original texts become, and the definition of translation comes to encompass forms such as adaptation, moving beyond the more orthodox view of translation as a faithful reproduction of the source text. This indeed seems to have been the case in Japan, in relation to source texts from both China and the West. Kambun kundoku - Making Chinese texts accessible to Japanese
readers
In early times the Japanese people did not have a script for writing their language. In the fifth century Chinese characters were adopted, although the two languages are totally unrelated, and in the eighth century an in genious system of diacritics and glosses was developed to represent the Japanese language by means of Chinese characters.2 This same annota tion method also enabled them to decipher original Chinese texts. This was possible because Chinese characters represented not only sound, but also meaning. Numbers and diacritics (variously referred to as transla tion marks, reading marks and transposition marks in the literature) were used to the left of the characters to show how to rearrange the Chinese word order to correspond to Japanese. Another kind of mark (kana) was used on the right to indicate the Japanese word inflections and grammati cal particles. The vocabulary and morphology reflected the language of the ninth century aristocratic class of Japan, with these speech patterns being imposed on Chinese to produce a hybridized Sino-Japanese that retained strong Chinese overtones, but which would “make little sense to anyone attempting to read it as Chinese” (Morris 1983:26). The result was “a partial as well as literal translation” (Backus ibid: 124). (This is somewhat akin to interlinear translations, where the aim is to show the basic meaning, word order and parts of speech, but no attempt is made to achieve a smooth rendition or to convey the fine nuances.) This transpo sition method, known as kambun kundoku (literally ‘Japanese reading of Chinese texts’) did not involve translating in the conventional sense, but instead used indicators to render Chinese texts comprehensible to Japanese readers. This practice changed over time and there were many variants. H. 2 By around the ninth century the Japanese had developed their own phonetic scripts, but they continued to use Chinese characters, partly because of the associ ated prestige and partly because they were more concise than works written in the phonetic script (Twine 1991:38). Chinese remained established as the written lan guage of scholarship while the Japanese language was used for writing poetry and popular literature. This combination of two distinct literary styles contributed greatly to enriching Japanese literature.
Wakabayashi: Marginal Forms of Translation in Japan
59
Sato (1994:11) writes that “some kundoku in the earliest periods was more like what we normally think of as “translation” as the “readers” made efforts to find indigenous words for Chinese expressions ...” (Note, however, that the burden lies with the reader, not the translator.) When reading aloud readers would read each segment in its Chinese pronun ciation, then again in its Japanese pronunciation, giving a fairly free colloquial rendition. By about the fifteenth century, however, the prac tice became more standardized and the renditions became much less free, with texts being pronounced in a Japanese approximation of the Chinese and Chinese expressions and syntax often being retained. This ultimate form of translatese assumes that readers are acquainted with the source language (H. Sato 1994:11). It was a direct means of construing a Chinese text into Japanese through a process of mental translation, how ever inaccurate the pronunciation, rather than reading it as a foreign language (Twine 1991:39). It required close attention to the original text, and this contributed to the literal approach to translation that has often been favoured in Japan. In the seventeenth and eighteenth centuries texts were sometimes written out in Japanese word order without using read ing marks, and both Chinese characters and the phonetic script developed in Japan were used. This was known as ‘direct translation style’, and was used well into the Meiji period (1868-1912) (Twine 1991:34-5). Thus kambun kundoku represents an instance of what Woods worth (1994:60) refers to as code-mixing, which occurs when a minority lan guage or literature (Japanese) is dominated by a major language or literature (Chinese). Texts written in the resulting ‘interlanguage’ (SinoJapanese) demand that their readers also be plurilingual, at once capable of reading and translating. Thus translation becomes an integral part of the reading experience. Opinions are divided as to whether or not kambun kundoku consti tutes translation. The term ‘translation’ is frequently used in reference to this, but is nearly always qualified in some way, with quotation marks of ten indicating its atypical status. Descriptions of this practice also indicate that it is both more and less than ‘real’ translation - that it is a “method of simultaneous reading and translating” (T. Sato 1983:194), but that the tone is not conveyed, and there are meaning discrepancies in the minor details (Kato and Maruyama 1991:351). Even when agreeing that kam bun kundoku does constitute translation, some writers maintain that its function is not to convert Chinese into Japanese, but to enable Japanese readers to grasp the gist of a text. Other writers (e.g. Kitamura 1993:13) argue that it is not translation, but a reading in the Japanized way which “does away with the need, in the true sense of translating the Chinese; it is rather a process of technical assimilation”. Indeed, discussions of trans lation in Japan often do not discuss kambun kundoku at all. Despite the
60
Unity in Diversity
centuries of contact with China and wide-ranging rendition of Chinese texts into kambun kundoku, translation ‘proper’ does not seem to be re garded as having started until the translation from Western languages that occurred from the sixteenth century onwards.
Adaptations of foreign works The practice of adapting foreign literary works has been very prevalent in Japanese literary history. In the Meiji period there were many free adaptations and summary-translations of western works, but as far back as the ninth century, traditional Chinese tales were rewritten in Japan. Different definitions of adaptation reveal the status of this practice in Japan in relation to translation.3 Yoshitake (1974:1,35) defines adap tation as a process whereby the style, expressions, structure, ways of thought and literary views of a foreign literary work are incorporated into the adaptor’s artistic repertoire and used as the basis for creating a new work. Unlike translations, adaptations are original creations, de spite being derived from another text. He includes both abridged versions and adaptations within the notion of translation (Yoshitake 1959:i-ii). Nogami (1938:225) takes a somewhat different approach, defining adaptation as the practice of borrowing the structure or thought of a text and rework-ing it to suit Japanese customs. This implies that the em phasis is on readers’ needs, rather than on fidelity. He notes that taking adaptation further leads to imitation, where the source text is no more than a kind of model, and argues that adaptations and imitations should not be regarded as translations. A further definition is given by M. Sato (1994:291), who uses the term ‘re-telling’ to refer to the practice of keep ing in mind an existing text when writing a new text. She includes in this category all texts except full translations, even where the source text is respected to a considerable extent. One factor behind the prevalence of adaptations in the late Edo period (1603-1868) was the attempt to avoid restrictions placed on the transla tion of western works. Another factor was that prior to the development of a modern written style, nineteenth century translators used a style that produced a discrepancy between the characters and their language, so that it seemed that westerners in Japanese dress were speaking. To over come this, adaptations in which the characters, settings and sentiment were all domesticated were used. Writers also used adaptations to hone their rhetorical skills. Korpel (1993:58) points out that in such cases fi delity to the original is regarded as ‘slavish’. 3The Japanese words for ‘translation’ (hon'yaku,) and ‘adaptation’ (hon’an) share a common element, the character hon, which means ‘turn, turn over, reverse’.
Wakabayashi: Marginal Forms of Translation in Japan
61
Sometimes the adaptations were closely modelled on the originals, but often they borrowed just the plot with varying degrees of fidelity. The first adaptation to use the plot of a Western novel was a 1774 work that was based on Gulliver's Travels and incorporated elements of a Ja panese legend and Chinese legends to give it colour. The focus was on making the plot interesting, rather than on literary value, and this ap proach was typical of the first stage of adaptations in Japan (Yoshitake 1974:3ff). Some adaptations add incidents to the original plot, such as Yamada Bimyo’s 1885 adaptation Tate goto Soshi of the legend of King Alfred, while others omit parts, such as Wakarejimo, Higuchi Ichiyo’s 1892 adaptation of Romeo and Juliet (Yoshitake 1974:142ff). Frequently, foreign themes or objects are adapted to the Japanese en vironment. For instance, some works by Takizawa Bakin (1767-1848) adapt Chinese themes to the Japanese setting. In 1895, the writer Tsubouchi Shoyo argued that in an adaptation the objects, people and even the emo tions should be ‘Japanized’ (Tomita 1965:162). Ozaki Koyo (1867-1903) transplanted tales from The Decameron into the atmosphere of Edo samu rai society, completely eliminating all western overtones and rewriting them in his own style. Yoshitake (1974:92ff) suggests that this is typical of the second stage of adaptations in Japan. From Ozaki onwards, adap tations gradually came to be regarded as literary works in their own right. The adaptations by Mori Ogai and Kitamura Tokoku represent the third stage of adaptations, in which the world view of the writer is incor porated (Yoshitake 1974:107ff). Kitamura Tokoku’s Horaikyoku (1891) resembles Byron’s Manfred (1817) but there is a fundamental difference in their views on love, and also differences in their views on religion (Yoshitake 1974:131). Another type of adaptation involves a move from one genre to another, such as prose renditions of poetry. On the outermost periphery were adaptations in which just one epi sode from the original work was incorporated in the Japanese work; works which used elements from numerous source texts; and ones in which the literary techniques of a particular author were imitated. Yet the position of all forms of adaptation is only peripheral in terms of conventional translation, and not necessarily in terms of literary significance. In fact, Keene (1987:71) writes that “Adaptations ... of European literary works, as opposed to translations, were of perhaps even greater importance in the development of modern Japanese literature”. Supporters of this prac tice (e.g. the literary magazine Waseda Bungaku) believed that the naturalization involved makes adaptations creative works that surpass translations in value, and that adaptations are equal to or nearly equal to original works (Tomita 1965:165). There have even been adaptations which have been regarded as superior to the original. Nevertheless, some observers (such as the magazines Taiyo and
62
Unity in Diversity
Teikoku Bungaku) believed that adaptations were a traitor to literature. At the end of the nineteenth century, adaptations were regarded as a poor second to translation, and Tomita (1965:163) claims that this disapproval was a reaction against their popularity. M. Sato (1994:290) points out that there has been considerable criticism of adaptations in children’s literature, and she describes the rivalry in children’s literature between full translations and adaptations, and the changing trends in this area since the 1950s. It seems, therefore, that not only is a distinction made in Japan between translation and adaptation, but also that they are ranked differ ently in terms of their value as creative works.
Conclusions Rather than taking a prescriptive approach, I have opted instead to simply describe the attitudes that have prevailed in Japan towards these two marginal practices. Although they do not constitute prototypical translation, they have both been carried out extensively in Japan and are significant in that they have been accorded a considerable degree of acceptance. Yet the ongoing disagreement over whether or not they comprise translation is indicative of the fuzziness of this concept in Japan. Going back to Toury’s definition, it seems that in Japan there has not been total acceptance of either kambun kundoku or adaptations as translation. The division be tween ‘proper’ translation and marginal practices remains, but is far from clearcut. Nevertheless, practices such as kambun kundoku and adapta tions, which transgress the conceptual borderlines conventionally applied to translation, may broaden our understanding of translation in general and provide one more piece in the overall picture of the “distinction be tween what is universal and what is culture, or language-specific” (Toury 1995:73).
References Backus, R.L. (1983) ‘Kambun’, in Kodansha Encyclopedia of Japan, Vol. 4, Tokyo: Kodansha Ltd, 123-124. Even-Zohar, I. (1978) Papers in Historical Poetics, Tel Aviv: The Porter Institute for Poetics and Semiotics, Tel Aviv University. Honma, H. (1974) Foreword to Y. Yoshitake’s Kindai Bungaku no Naka no Seio - Kindai Nihon Hon’an-shi, Tokyo: Kyoiku Shuppan Center, 1st edition. Kato, S. and Maruyama, M. (1991) Hon’yaku no Shiso - Nihon Kindai Shiso Taikei 15, Tokyo: Iwanami Shoten. Keene, D. (1987) Dawn to the West: Japanese Literature in the Modern Era, New York: Henry Holt & Company.
Wakabayashi: Marginal Forms of Translation in Japan
63
Kitamura, I. (1993) ‘Problems of the Translation of Law in Japan’, Victoria University of Wellington Law Review Monograph 7. Korpel, L. (1993) ‘Rhetoric and Dutch Translation Theory (1750-1820)’, Target 5(1): 55-69. Morris, M. (1983) ‘Early literature’, in Kodansha Encyclopedia of Japan, Vol. 5, Tokyo: Kodansha Ltd, 25-28. Nogami, T. (1938) ‘Honyaku no Taido’, in Y. Kawamori (ed) Flon’yaku Bungaku- Kindai Bungaku Kanshö Koza 21, 1961, Tokyo: Kadokawa Shoten, 224-234. Pedersen, V. H. (1995) Book review of M. Schreiber, Perspectives: Studies in Translatology 1: 139-140. Sato, H. (1994) ‘Using a Foreign Language’, JLD Times, April 1994, 1 and 10- 12. Sato, M. (1994) ‘Jidö bungaku no saiwa - sono haikei to kanosei’, in S. Kamei (ed) Kindai Nihon no Hon’yaku Bunka, Tokyo: Chuo Koronsha, 289-306. Sato, T. (1983) ‘Poetry and Prose in Chinese’, in Kodansha Encyclopedia of Japan, vol. 6. Tokyo: Kodansha Ltd, 193-197. Tomita, H. (1965) ‘Meiji chuki no hon’yaku oyobi hon’an ron’, Hikaku bungaku nenshi, January 1965, 149-174. Toury, G. (1980) In Search of a Theory of Translation, Tel Aviv: Porter Institute. Toury, G. (1995) Descriptive Translation Studies and Beyond, Amsterdam/ Philadelphia: John Benjamins. Twine, N. (1991) Language and the Modern State: The Reform of Written Japanese, London & New York: Routledge. Woodsworth, J. (1994) ‘Translators and the Emergence of National Litera tures’, in M. Snell-Homby, F. Pöchhacker and K. Kaindl (eds) Translation Studies: An Interdiscipline, Amsterdam/Philadelphia: John Benjamins, 55-63. Yoshitake, Y. (1974) Kindai Bungaku no Naka no Seid - Kindai Nihon Hon’an-shi, Tokyo: Kyoiku Shuppan Center, 1st edition.
Page Intentionally Left Blank
Choices and Constraints in Screen Translation EITHNE O’CONNELL SALIS, Dublin City University, Ireland The two most common forms of screen translation are dubbing and subtitling and, historically, most countries have tended to adopt one or other of these two language versioning methods to the virtual exclusion of any other. This article offers some possible explana tions for past trends and outlines likely reasons for the changing fortunes of both interlingual and intralingual subtitling within the EU. It also evaluates the relative suitability of a variety of types of screen translation in relation to such variables as audience age, sex, education and social class as well as programme genre and, of course, linguistic, economic and political considerations. Thanks to the opportunities offered by satellite and cable technology, improved film and video distribution methods and the latest screen trans lation techniques, film makers and broadcasters in Europe now have the opportunity to reach much wider and more heterogeneous audiences than was possible in even the recent past. This augurs well for an industry which has been in the doldrums for most of the century. In the very early days, European cinema was in an enviable position and “between 1906 and 1913 the output of the French industry alone accounted for over one third of global box office receipts” (Flynn 1995:15). But when war broke out in Europe, investment fell off and the US pushed ahead to achieve the virtually unassailable position in the audio-visual world it still holds today. One problem which continues to dog the European industry is the fact that although Europe has many more television viewers and cinema goers than North America, the European audio-visual market is fragment ed into many linguistic regions. Consequently, European film and TV programmes are produced in the context of a series of small and largely unconnected markets and are at an immediate disadvantage compared to US productions in terms of recouping their costs within their national territories (Watson 1992:14). Clearly, if the European audio-visual in dustry is to grow and prosper, it will have to learn to cross linguistic frontiers by using screen translation methods to reach larger audiences. The two most common forms of screen translation are dubbing and subtitling. Dubbing may be understood as a general term covering vari ous kinds of revoicing such as voice-over, narration and free commentary although it frequently refers more specifically to what is also known as lip-synch dubbing. Subtitling is usually associated with interlingual
66
Unity in Diversity
transfer, although the advent of Teletext technology, in particular, has contributed to the increasing importance of intralingual subtitles, aimed primarily at those with impaired hearing. Although frequently mention ed in the same breath, dubbing and subtitling are very different forms of screen translation. As Delabastita (1989:196) has pointed out, screen communication occurs through two channels rather than one, namely the visual channel and the acoustic channel. The visual channel can, of course, also be used to transmit verbal signs, e.g. the title of the film or programme, credits etc., while the acoustic channel can be used to pass on non-verbal as well as verbal information, e.g. the sound of a door opening. Essentially, dubbing relates to the acoustic channel as it involves the replacement of source acoustic verbal signs, while subtitling effects changes relating to the visual channel, retaining the source acoustic ver bal signs but adding target language visual verbal signs as well. In short, dubbing is a process of acoustic replacement while subtitling is a process of visual supplementation. The ways in which these two processes are carried out are so different as to hardly warrant comparison. In the case of dubbing, the final product is very much the result of a collaborative effort involving the original screenplay writer, the translator, dubbing actors, the dubbing director, sound studio technicians, and so on. While the precise nature of the task involved depends on the work to be dubbed, the process is usually a highly complex, lengthy and consequently expensive one. The primary constraint imposed on dubbing relates to what Fodor (1976) called phonetic syn chrony (now generally known as the problem of lip-synch) although content and character synchrony can also be problematic. Ideally, the process involves the replacement of the original source language voicetrack by a translated version which attempts to be as faithful as possible both in terms of reproduction of the semantic content as well as the tim ing, phrasing and lip movements of the original (Luyken 1991:73). According to Delabastita, some commentators exaggerate the signific ance of lip-synch as a constraint in dubbing. While it is true that close-up shots of a character speaking can pose such problems for lip-synch that a phrase or sentence has to be altered fundamentally in translation, it should be remembered that: on the other hand, in many scenes the character who speaks is not even within view. The angle and distance of the camera and the general visibility conditions are important factors here. So is the narrative structure of the film; the use of an off-screen narrator in the original film, for instance, drastically simplifies matters from the point of view of synchronisation. (1989:203)
O'Connell: Choices and Constraints in Screen Translation
67
Subtitling is a very different kind of screen translation. While excep tional circumstances may necessitate that a particular piece of work be divided up, there is normally no particular reason why more than one person should be involved in the process and, as a consequence, subti tling tends to be much less expensive than dubbing. Since people generally speak much faster than they read, subtitling inevitably involves textual constraints such as the need to reduce the original message as well as the technical constraints of shortage of screen space and lack of time. The subtitler endeavours to retain as much as possible of the original but has the usual translation problems of transfer between two languages and cultures compounded by the specific constraints of this mode of screen translation. For much of the history of film and TV, subtitling was viewed as a poor second to dubbing. But now all that seems to be changing. The in creasing popularity of subtitles is certainly helped by the relatively low costs involved but another very significant factor is the growing interest many Europeans now have in their neighbours, and their cultures and languages. As Danan (1991:613) has observed, subtitling “indirectly pro motes the use of a foreign language as an everyday function in addition to creating an interest in a foreign culture”. In short, subtitling for all its imperfections amounts to an inexpensive, quick, foreign-culture friendly and generally fairly politically correct mode of screen translation. For these reasons, the EU through its MEDIA II programme is promoting this method for wider use throughout its member states. Papers on screen translation often start by pointing out that the coun tries of Western Europe can be subdivided into dubbing countries (e.g. France, Germany, Spain, Italy) and subtitling countries (e.g. Belgium, Denmark, Finland, Greece, Portugal, Sweden). Usually this division is explained in economic terms, i.e. dubbing can be up to ten times more expensive than subtitling and is therefore not an option for smaller coun tries (Gambier 1994:243). Sometimes historical or political explanations are offered as well. For example, the development of the European sound film industry in the 1930s coincided in many countries with a growth in nationalism, and the dubbing of films in particular offered an ideal op portunity to adapt original screenplays to suit the prevailing political philosophy of specific European countries at the time: A strong nationalistic system tends to be closed and reject or limit outside influences, since the home system is perceived as the embodiment of a firmly-established, superior tradition ... Translation in a nationalistic environment must therefore be target-orientated in order to make foreign material conform as much as possible to the local standards. In this sense, dubbing is target-orientated ... an
68
Unity in Diversity
attempt to hide the foreign nature of a film by creating the illusion the actors are speaking the viewer's language ... an assertion of the supremacy of the national language and its unchallenged political, economic and cultural power within the nation boundaries. (Danan 1991:612) The reason that dubbing lends itself so readily to manipulation for political purposes is that it represents a ‘covert’ mode of translation. Sub titling, on the other hand, is seen as ‘overt’ translation “laying itself bare to criticism from everyone with the slightest knowledge of the source language” (Gottlieb 1994:102). While discussions of economic and political issues in screen transla tion cast some light on the background to the current divisions along national lines between dubbing and subtitling countries, they fail to ex plain, for instance, why large Latin American countries as well as the tiny Basque Country prefer to dub while Romania subtitles virtually everything (Dries 1995:6). In reality, choices made in relation to dubbing vs. subtitling are often the result of the complex interaction of a number of factors in addition to the merely economic, political and/or historical. In other words, cultural, educational and linguistic considerations can also have a place in the equation (O’Connell 1996:152). In fact, any at tempt to provide a convincing explanation as to why a particular film or TV programme is, or should be, dubbed rather than subtitled (or vice versa) should consider making reference to target audience profile (e.g. age, sex, educational background, social class) as well as programme genre, cost, time, the status of the source and target languages (e.g. world, major, minority languages) and the power relations which exist between them (Cronin 1995:88). Neither should the explanation ignore such is sues as the primary intention of the broadcaster or film maker (e.g. entertainment, education, propaganda, art) and the primary intention of the person who commissions the language transfer method (e.g. to reach the largest possible audience, to develop LI or L2 competence or, very commonly nowadays, to facilitate those with impaired hearing) (O’Con nell 1994:371). Basque must surely qualify as one of Europe’s smaller minority lan guages. If the simplistic explanations which cite economic and historical/ political considerations as the key factors influencing the choice of screen translation method were accurate, it would follow that Basque TV would opt for subtitling, thereby saving up to 90% of the cost of dubbing the same programme. Yet despite the fact that small, neighbouring countries like Portugal tend to subtitle, the Basque Country does the opposite and dubs its programmes at considerable cost, as do its relatively wealthy neighbours, France and Spain. However, these countries broadcast in major
O'Connell: Choices and Constraints in Screen Translation
69
European, and indeed world, languages and have enormous audience potential compared to the Basque Country. The explanation is hardly that Basque TV wants to copy conventional practice in France and Spain regardless of the expense and, since Basque TV is a relatively new phe nomenon, the explanation cannot be historical either. In this case, the choice made by the Basque broadcasting authorities must be understood primarily in terms of language politics and language planning. Speakers of Basque, like speakers of Welsh and Irish, lead precarious linguistic lives in an environment in which many social institutions such as schools, courts and health services, function largely, if not exclusively, through another more widely spoken language, for example Spanish, French or English. Since one of the most striking features of subtitling is the fact that, unlike dubbing, it leaves the original soundtrack intact, it follows that speakers of lesser-used languages would be ill-served if foreign language productions were subtitled rather than dubbed into their languages as they would, in effect, end up listening to a foreign soundtrack while try ing to read subtitles in their own language. Such a scenario would have the effect of further undermining already weak linguistic communities by exposing them unnecessarily to major languages on an aural level while restricting audio-visual communication in the lesser-used languages to the written code (O’Connell 1994:372). This explains why countries like Wales, Ireland and the Basque Country feature on the list of dub bing enthusiasts. But when one looks more closely, one becomes aware that this choice of screen translation method does not apply right across the board in these countries since even here factors other than language planning considerations may also come into play and place certain con straints on the original choice of dubbing. W ales, for example, is commonly regarded as one of the dubbing countries. However, second ary considerations, such as the wish to reach Anglophone viewers and to cater for the needs of those with impaired hearing, have resulted in the provision of English Teletext subtitles on many programmes. Imported children's programmes such as cartoons, if aimed at those under 8 years of age, clearly need to be dubbed in any country regardless of whether or not the usual broadcasting convention is to use subtitles. However, broad casters at S4C, the Welsh-language TV channel, have found that it is necessary, for language planning reasons, to subtitle playschool-type pro grammes into English so that Anglophone parents, who are having their children educated through Welsh, can understand the material their chil dren are viewing. Quite apart from the special case of young viewers, the age of the main target audience can be a significant factor placing various constraints on the choice of screen translation method. The extensive use of Teletext
70
Unity in Diversity
subtitles nowadays is largely due to the effective lobbying work of groups representing those suffering from impaired hearing. Yet a subset of these people experience hearing difficulties due to advanced years and may also have difficulties reading the subtitles because of poor eyesight. In other words, a reasonably convincing case can be made for or against the choice of either dubbing or subtitling by citing the factor ‘age’. The sex of viewers might not immediately spring to mind as a crucial factor which could influence the choice of language transfer method but evidence has emerged in recent years which suggests that certain differ ences in UK viewing preferences can be discerned between the sexes. Research over the last decade shows that the British in general are still reluctant to watch dubbed or subtitled programmes but young, well edu cated males apparently enjoy subtitled material (Kilborn 1989:430). Perhaps this has something to do with their foreign language skills and general interest in other cultures. On the other hand, women and working class men generally prefer imported programmes to be dubbed. In the case of the former, the fact that women often combine TV viewing with domestic chores may well play some part in the explanation while in the case of the latter, a certain lack of familiarity with foreign cultures and, perhaps, literacy problems may be the key. Although these examples were cited to illustrate the different viewing preferences of men and women, they also show that the educational back ground of a particular target audience can act as a constraint in screen translation. Elsewhere, as in Latin American countries, factors such as cost (which in Europe can be of paramount importance) are greatly out weighed by this question of the audience’s educational background. If literacy levels are low, there is clearly no real option but to dub for televi sion and/or cinema regardless of cost. In conclusion, it is worth mentioning some interesting findings relat ing to constraints and choices in screen translation which have emerged from research conducted by the European Institute of the Media in rela tion to audience preference. “On the whole, the proportion of the choice between the main two forms of Language Transfer ranged from 20 to 1 in West Germany to 5 to 1 in Switzerland and was always in favour of the method to which the audience in question was accustomed” (Luyken 1991:112). It would therefore appear to be the case that past and present preferences for particular modes of screen translation will not necessar ily act as a major constraint on the choices to be made in the future. Since the evidence shows that current audience preferences are “determined by familiarity and conditioning”, it seems likely that they are not “unal terable and they might be transformed by familiarisation with other alternatives” (ibid). This can only be good news for the European audio visual industry in general which hopes to win back more and more
O'Connell: Choices and Constraints in Screen Translation
71
markets from the Americans. But it is particularly good news for advo cates of subtitling, the screen translation method which was for so long the underdog but now appears to have realistic chances of being much more widely used throughout the EU.
References Cronin, M. (1995) ‘Altered States: Translation and Minority Languages’, TTR VIII(l): 85-103. Danan, M. (1991) ‘Dubbing as an Expression of Nationalism’, Meta, XXXVI(4): 606-613. Delabastita, D. (1989) ‘Translation and Mass Communication: Film and T.V. Translation as Evidence of Cultural Dynamics’, Babel XXXV( 4): 193-218. Dries, J. (1995) ‘Breaking Eastern European Barriers’, Sequentia 11(4): 4. Fodor, I. (1976) Film Dubbing: Phonetic, Semiotic, Esthetic and Psychologi cal Aspects, Hamburg: Buske. Flynn, R. (1995) ‘Europa, Europa: Ireland and Media II’, Film Ireland April/ May: 15-17. Gambier, Y. (1994) ‘Subtitling: a Type of Transfer’, in F. Eguiluz (ed.) Transvases Culturales: Literatura, Cine, Traducción, Vitoria: Facultad de Filologia, 243-251. Gotttlieb, H. (1994) ‘Subtitling - Diagonal Translation’, Perspectives: Stud ies in Translation I: 101-121. Hesse-Quack, O. (1969) Der Übertragungsprozess bei der Synchronisation von Filmen. Eine interkulturelie Untersuchung, Munich/Basle: Reinhardt. Kilborn, R. (1989) ‘They Don’t Speak Proper English: A New Look at the Dubbing and Subtitling Debate’, Journal of Multilingual and Multicultural Development X(5): 421-434/ Luyken, G. (1991) Overcoming Language Barriers in Television, Manches ter: EIM. O’Connell, E. (1994) ‘Media Translation and Lesser-Used Languages’, in F. Eguiluz (ed) Transvases Culturales: Literatura, Cine, Traducción, Vitoria: Facultad de Filologia, 367-373. O’Connell, E. (1996) ‘Media Translation and Translation Studies’, in T. Hickey and J. Williams (eds) Language, Education and Society in a Chang ing World, Clevedon: Multilingual Matters, 151-156. Watson, N. (1992) European Film Feasibility Study, Madrid: Media Busi ness School, 14-17.
Page Intentionally Left Blank
Section Three Descriptive Translation Studies
Page Intentionally Left Blank
Six Subtitlers - Six Subtitling Texts IRENA KOVACIC Faculty of Arts University of Ljubljana, Slovenia In this experimental study, six Slovene subtitlers were asked to trans late a passage from a television drama (Eugene O’Neill’s Long Day’s Journey into Night). Using a framework based in part on Halliday's model of functional linguistics, the author compares the following textual parameters used by the six subtitlers: number and organization of subtitles; text reduction in terms of linguistic func tions; language registers; dramatic and conversational structure. The objective of the analysis is to determine the extent to which the subtitlers followed (either consciously or unconsciously) the same strategies and in what kinds of situation their translational solu tions differed most significantly.
Introduction Subtitling is a specific form of translating in which additional extralinguistic constraints have to be taken into consideration (limited space, synchronization with the image). Apart from that, it may be regarded in the same way as any other translation: its objective is to render a source language text into a form that will make its meaning potential accessible to a target audience (Hatim and Mason 1990:10-11). Even its constraints may be accounted for within a general framework of translation studies: subtitles are just a special text type and their form and the procedures involved in their production are dictated by their anticipated function, which is to facilitate reception and comprehension of a film or television programme produced in a foreign language. Empirical evidence shows unequivocally that different translators - even when working in very similar circumstances and environments will produce very different translation texts. Cognitive sciences and their reflections in literary theory offer an explanation for this: all transla tors are first of all readers of the original text and their reception of the original is more or less idiosyncratic, determined by their respective cog nitive environments. Translators’ external environments, the cultures in which they live, with their traditions and values, participate in the forma tion of their cognitive environments, but part of the latter is shaped by very individual experiences and influences (Iser 1978; Tompkins 1980). However hard translators may try, they will still not be ideal readers,
76
Unity in Diversity
but will succumb to at least some of the influences, typically those of which they are not aware and therefore cannot control. These collective and individual influences will eventually determine the translator’s ap proach to translating a text. A derivative of this concept of different readings of the same text is by now a well established description of trans lating as rewriting and translations as texts in their own right (Lefevere 1992:1-10).
The experiment Starting from the notion of translations as reflections of individual read ings and rewritings of the original text, in an experimental study six Slovene subtitlers were asked to translate a passage from a television drama (the 1986 Miles Company television adaptation of a Broadway production of Eugene O’Neill’s Long Day's Journey into Night, starring Jack Lemmon, Bethel Leslie, Peter Gallagher and Kevin Spacey). Their translations were analyzed in terms of various textual parameters. The six sets of subtitles were analyzed for the following textual features: • number and organization of subtitles, • text reduction in terms of linguistic functions, • language registers (including treatment of culture specific terms), • dramatic and conversational structure.
Number and organization of subtitles Slovene subtitlers split dialogues into subtitles themselves, as part of the translation process. The well-known 7-second principle (i.e. on av erage, it takes viewers six to seven seconds to read a two-line subtitle) is a general guideline, but individually adjusted to the dynamics of the dia logue. Of the six subtitlers, five used very similar numbers of subtitles (between 120 and 134, with an average of 129), while one text deviated considerably with its 157 subtitles; this gave a total average of 134. Subtitles may consist of one or two lines. In the experiment, there are only 4% of one-liners (32 out of 802). These are most frequent in texts produced by subtitlers D and E (8.9% and 6.7% respectively) - for quite different reasons: as subtitler D reduced the text less than the others (157 subtitles), she occasionally needed an extra line to complete a two-liner; subtitler E made the most reductions (120 subtitles), which means that at certain places one line was sufficient for him where the others needed two lines. A second general difference that can be observed in the organization of subtitles is between those that end with a high-rank separator indicat ing sentence boundaries (a full stop, question mark, exclamation mark or suspension dots) and those cut off at other places, either with a lower-
Kovacic: Six Subtitlers Six Subtitling Texts -
77
rank punctuation mark or with no punctuation at all. The percentage of subtitles ending with sentential punctuation marks ranges from 76.7% (subtitler F) to 97.8% (subtitler A). According to what we know about the reading process, cutting off subtitles at sentential borders seems pref erable, because sentential borders are the usual spot where our eyes stop during reading to wrap up the content of the last stretch of the text. Some subtitlers, however, insist that it is better to continue a sentence into the next subtitle in order to ensure continuity. Even in this case, the subtitlers in the experiment generally observe the next lower grammatical unit, and cut off subtitles at clausal boundaries. Of the two subtitlers who used non-sentential cut-offs, subtitler F (23.3% of non-sentential cut-offs) con sistently used this principle to follow the rhythm of the dialogue (see also below), while subtitler D ended up with 11.4% of non-sentential cut-offs probably because she had reduced the text less than the others. Conse quently, the space available in a subtitle was simply not sufficient and she had to carry over into the next one. A third visible difference in the arrangement of subtitles is their rela tion to conversational organization: they can contain utterances by one or more characters. In the material analyzed, no one-liner includes a turntaking point. Also among the two-liners, one-speaker subtitles prevail (72.2 vs. 27.8%). The discrepancy is smaller here than in the previous categories, with the percentage ranging from 68.2 (subtitler F) to 75.5 (subtitler D). A lower number of two - (or more) - speaker subtitles may be due to two different subtitling strategies. The first one involves treat ing of such elements of conversation as back-channel signals or short interruptions (see below). If a subtitler prefers to ignore them and ‘smooth out’ the dialogue, the resultant text will contain fewer more-speaker sub titles. The other reason is treatment of turn-taking points. While some subtitlers tend to treat one character’s turn as the most tightly knit unit and make their subtitles run parallel with turns, others want to emphasize the interactive nature of conversation by bringing together two or more characters in one subtitle. In the material analyzed, this is most obvious in the case of subtitler F, who used the highest percentage of two- (or more-) speaker subtitles (31.8%) - but also the highest percentage of non-sentential cut-off points. The two together imply a consistent strat egy to achieve an impression of continuity or fluid progression throughout a story by using signals of linguistic and conversational incompleteness. At the other extreme is subtitler D with 24.5% of two- (or more-) speaker subtitles and 75.5% of one-speaker subtitles. This result complements the finding that she used more one-liners than the others. Her prevailing strategy (besides conservative reductions) appears to be that a new turn demands a new subtitle, regardless of whether there is enough material for a two-liner or not.
78
Unity in Diversity
Text reduction in terms of linguistic functions A key question in any analysis of subtitles is which parts of the original text get translated, which are translated only partially, in a condensed form, and which are completely omitted. In the study, only omissions were examined in more detail. In this section of the analysis, Halliday’s (1985a) model of linguistic functions, adapted to incorporate elements of conversational interaction, was used. An earlier study (Kovacic 1991; Kovacic 1992) showed that it may be reasonable to assume that the function performed by a linguistic element (at any level of language or text structure) determines signifi cantly how that element will be treated in subtitles, i.e. whether it will be preserved, reduced or discarded. Ideational elements tend to be preserv ed most; when they are omitted, the reasons may be found in the semantic structure of the text. Personal and interpersonal elements are frequently omitted because they are redundant when combined with pictures and also because they fall into the category of typical elements of spoken language which do not always readily find a written realization. Textual elements are comparatively less important than, for example, ideational elements for two reasons: firstly, because coherence of dialogue is sup ported by continuity of visual material and secondly, because in fictional dialogues the course of a conversation is not really managed by the par ticipants but by the screen writer as its creator (as it is in real life, with textual elements playing a significant role in the process). In the experiment, the findings of Kovacic 1992 were confirmed. Pro portionally, elements with an ideational function were omitted less frequently than (inter-) personal and textual elements. The difference between the latter two was not statistically significant. In some catego ries (e.g. omission of vocatives), the subtitlers reached a very high degree of agreement, while in others (e.g. omission of modality) they differed considerably. While the lowest number of subtitles correlated with the highest number of omissions (subtitler E) and the highest number of sub titles correlated with the lowest number of omissions (subtitler D), the other relations were not similarly systematic. For example, in text B (130 subtitles), 212 omissions were identified, while translator C in her 132 subtitles omitted as many as 298 elements (the second highest number). The explanation is twofold: (a) subtitler B was very efficient in finding more economical renderings and skilful in the complementary reduction procedure, i.e. condensation, and (b) subtitler C used a more literary, figurative language, with more metaphors and other figures of speech, which, however, takes up more space. In other words, text B lost less in comparison with the original, while text C was more vivid (sometimes even excessively figurative - see below). As to subtitler C, a reason for
Kovacic: Six Subtitlers - Six Subtitling Texts
79
her economy and efficiency may be sought in the fact that she is the only person in the experimental group to hold a degree in English and Slovene (all the others have graduated in English and some other subject). Text D, which contains the highest number of subtitles and the lowest number of omissions, does not deviate significantly from the others in its omission of (inter-) personal and textual elements. The difference occurs in the ideational component. Translator D preserved more than the others elements such as attributes, circumstances, co-ordinated and appositional constructions, information expansions, etc. These are the usual candidates for omission in the ideational sphere, but no general rule can be postulat ed. If they denote a feature or circumstance that is not indispensable for the understanding of the dialogue or the plot (e.g. some attributes or cir cumstances), if they can be deduced from the context or our general knowledge (e.g. some co-ordinated constructions or parts of expansions), they may be left out, assuming that the viewer will apply the principle of relevance to fill in the missing information. Such reductions (both omis sions and condensations) make a text shorter, but more informative. In terms of the word/information ratio, viewers have to invest more energy in transforming the words they are given into an image of the fictional world. Part of the subtitler’s translation strategy is to find the right bal ance between economy and comprehensibility. Too much text may be impossible to read and the viewer will not be able to follow the story. Excessive reduction will either demand from viewers additional effort and time to fill in the missing parts or will leave them at a loss as to what is going on in the story. Different viewers will need different amounts of redundant and complementary information, depending on their back ground knowledge, decoding procedures, etc. The decision as to what is necessary and what is superfluous in translation is very individual and will differ from subtitler to subtitler. In our experiment, subtitler D differ ed from her colleagues in this respect, preserving in her translation more ideational information, sometimes even at the expense of readability.
Language registers Subtitling is associated with a general problem of register. Subtitles are written, yet they are supposed to reflect spoken communication. Further more, it is well known (cf. Halliday 1985b) that spoken and written language have a number of distinctive features, that written language cannot incorporate all of the information conveyed by speech (suprasegmental features) and that some typical features of spoken language appear definitely odd and distracting in written form (hesitations, incom plete utterances, interpersonal signals). This raises a theoretical question regarding the status of subtitles: are they an independent stratum of
80
Unity in Diversity
language, intermediate between speech and writing, or are they rather a sub-category of either of the two (cf. also Lambert 1993:234). In coun tries like Slovenia, with a strong prescriptive tradition, subtitles have long observed almost exclusively norms of written language (under the influ ence of the big brother, literary translation, and the status of television as a mass media which should contribute to improving the general level of language use). It is only with the arrival of younger subtitlers and their changed understanding of language functions that the concept of lan guage variability has also penetrated into subtitles. The process is far from complete, however, and no proper linguistic standards for subtitles exist. As a consequence, subtitlers exercise their own judgement as to the appropriate register or style, with some influence from TV compa nies’ language editors. The six subtitlers in the experiment differed considerably in their se lection of language registers. The original did not contain slang or vulgar expressions, which are often a delicate issue in subtitling. The main prob lem for the subtitlers was how to choose between more literary and more conversational options. Of the six texts, text C displays the highest num ber of literary features (metaphors and other figures of speech, literary expressions). The impression of literariness is further intensified by the highest number of omissions of elements with (inter-) personal function (including exclamations, reinforcing formulas, and vocatives) and dis ruptive interventions in conversation, with only moderate omission of textual elements. Subtitlers D and E came closest to a conversational style, mainly through their selection of vocabulary, including replacement of literary metaphors by those common in everyday language. However, they too did not attempt to overstep the limits of standard grammatical correctness. A viewer who would not follow the original text but only the subtitles might be left with different impressions about the social status of the family in the play or the stylistic intentions of the author.
Dramatic and conversational structure In a text like Long Day's Journey into Night, the characters’ utterances do not have a merely communicative function; they also serve to create a certain atmosphere in the play, to characterize persons and relations among them. Repetition of certain expressions, re-use of adjectives, phrases or exclamations that may appear unimportant in their immedi ate context, play a very different role when viewed from the point of view of the entire text. In the text under analysis, such a role is performed by repeated reference to fog, mother’s constant anxiety about her appear ance, her complaints that she cannot see well any more. A translator sensitive to the dramatic pattern of a text will try to be consistent in pre
Kovacic: Six Subtitlers - Six Subtitling Texts
81
serving the small details that contribute to the overall atmosphere, while another who focuses on immediate linguistic expressions and sentence meaning may overlook the relevance of such elements. In the latter case, the subtitles will read as rather purposeless conversation, without show ing the tense relations underneath these seemingly casual exchanges. Another feature of conversation which signals the neurotic relations among family members is their eagerness to cut off conversation before it develops into conflict, and to cover for tension by frequent back-channel signals of agreement or by seemingly playful, teasing interruptions. From the point of view of sentential information, such interruptions are only disruptive and disturb the viewer in following the main thread of a con versation. That is why they are frequently omitted in subtitles. However, when subtitlers are attentive to the psychological value of interrup tions, they will prefer to sacrifice some ideational content and record the interruption. Similarly indicative of a person’s state of mind are hesitations, false starts, incomplete sentences and other similar features of spoken lan guage. A translator who regards subtitles as a subgenre of written literary works will consistently tend to reduce these features to a minimum. In the experiment, even the translators who attempted to simulate conversa tional style rarely used such syntactically irregular or incomplete patterns. A possible reason for this may be that, in the original, they typically occur in very rapid and overlapping exchanges; these were radically reduced in the subtitles.
Conclusion When all these aspects are put together, the result is a complex (yet far from complete) network of relations: fewer subtitles - more omissions (but not distributed equally across the different linguistic functions), a more literary style - fewer (inter)personal elements, greater awareness of the dramatic structure of a text - more attention to details, etc. All this is on the level of language alone. Other parameters would appear if the analysis also included interaction of language and picture. Very little is known at the present stage about the interaction of subtitles and picture in the viewers’ reception of television programmes. It would be interesting to study what (if any) influence these six different sets of subtitles might have on viewers’ understanding and reception of the play. Would it still be more or less the same play? If yes, this would mean either (a) that language plays a relatively subordinate role in television programmes, or (b) that the differences observed in the six sets of subtitles belong to some non-core domains of language and do not significantly influence the overall impact produced by the text. Another dimension of research
82
Unity in Diversity
might be added by comparing subtitles of the same programme into several languages and then studying the strategies used in translation, solutions of specific situations, or problems encountered in one language (country, society) and non-existent in another.
References Halliday, M. A. K. (1985a) Introduction to Functional Grammar, London: Edward Arnold. Halliday, M. A. K. (1985b) Spoken and Written Language, Oxford: Oxford University Press. Hatim, B. and Mason, I. (1990) Discourse and the Translator, London: Longman. Iser, W. (1978) The Act of Reading: A Theory of Aesthetic Response, Balti more: Johns Hopkins University Press. Kovacic, I. (1991) ‘Subtitling and Contemporary Linguistic Theories’, in M. Jovanovic (ed) Translation, A Creative Profession. Proceedings of the XII World Congress of FIT, Beograd 1990, Beograd: Prevodilac, 407417. Kovacic, I. (1992) Jezikoslovni pogled na podnaslovno prevajanje televizijskih oddaj (Linguistic Aspects of Subtitling Television Programmes), PhD thesis, Ljubljana: University of Ljubljana, Faculty of Arts. Lambert, J. (1993) ‘Le sous-titrage et la question des traductions. Rapport sur enquête’, in R. Amtz and G. Thome (eds) Übersetzungswissenschaft. Ergebnisse und Perspektiven, Tübingen: Gunter Narr, 228-238. Lefevere, A. (1992) Translation, Rewriting, and the Manipulation of Liter. ary Fame, London: Routledge. O’Neill, E. (1956) Long Day’s Journey into Night, New Haven & London: Yale University Press. Tompkins, J. (ed) (1980) Reader-Response Criticism. From Formalism to Post-Structuralism, Baltimore: The John Hopkins University Press.
Parallel Texts in Translation CHRISTINA SCHAFFNER Department of Languages and European Studies Aston University, Birmingham, UK Creating an appropriate translation often means adapting the tar get text (TT) to the text-typological conventions of the target culture. Such knowledge can be gained by a comparative analysis of paral lel texts, i.e. L2 and LI texts of equal informativity which have been produced in similar communicative situations. The author discusses some problems related to (cross-cultural) text-typological conven tions and the role of parallel texts for describing translation strategies. Implications for teaching translation are also discussed. The discussion is supported with examples of parallel texts that are representative of various genres, such as instruction manuals, in ternational treaties, and tourist brochures.
Introduction Everyone working in the vast area of translation studies is familiar with different approaches or models used to describe and explain the phenom enon of translation and interpreting. Neubert and Shreve (1992:12ff.) list the critical, the practical, the linguistic, the text-linguistic, the sociocul tural, the computational and the psycholinguistic models. Another aspect of diversity concerns the choice (or deliberate avoidance) of specific terms a case in point being ‘equivalence’ (cf. Snell-Hornby 1988:13ff.). One candidate for unity is the interest in the quality of the target text (TT) to be produced. In contrast to other instances of text production, translation is a case of secondary text production, or a ‘source-textinduced target-text-production’ (Neubert 1985:18). In translation, we are always dealing with texts-in-situation and in culture. Thus, the source text (ST) was produced (and received) in a particular situation (time and place) with a particular purpose for addressees in the source culture who have (culture-) specific knowledge, experience and expectations. In ad dition, the ST is an exemplar of a text type, or genre, e.g. an instruction manual, a news report, a fairy tale. The addressees’ knowledge includes knowledge of such text types. The ST, thus, has fulfilled its intended function in the primary communicative situation in the source culture. As a result of a translation process, a TT is produced that functions in a secondary communicative situation for target addressees with their re spective knowledge and expectations in a target culture. This can schematically be represented as shown in figure 1:
Unity in Diversity
84 Figure 1. Schema of translation addressees in SC (with experience, expectations, knowledge)
addressees in TC (with experience, expectations, knowledge)
ST
i
situation purpose text type primary communicative situation in SC
i
situation purpose text type secondary communicative situation in TC
SC = source culture; TC = target culture; ST = source text; TT= target text; TR = translation process/translator The inclusion of the text type points to the consideration of text-typological conventions as a criterion for the quality of the TT. Therefore,
we could say that a TT is of high quality when it conforms to the texttypological conventions of the target culture, i.e. conventions the TT addressees are familiar with and would expect in the specific situation. Knowledge about text-typological conventions can be gained from a sys tematic analysis of parallel texts1 which are ‘L2 and texts of equal informativity which have been produced in more or less identical com municative situations’ (Neubert 1985:75). The comparative description of text types is a fairly recent development, although as early as 1980 Hartmann called for a ‘contrastive textology’. In the domain of textlinguistics, and specifically LSP research (Language for Special Purposes), the aim has more often been the description of text types in one language (e.g. Lenk 1993; Gläser 1990) rather than from a comparative perspec tive. In translation studies, the relevance of parallel texts has been widely recognized, but the information gained from such studies has been mainly selective. 1The use of the term parallel here differs from that of some corpus linguists (see Laviosa and Peters and Picchi, this volume), who use the term to refer to a corpus of source texts and their translations into a target language.
Schaffner: Parallel Texts in Translation
85
Conventionalized text types The relevance and advantages of a translation-oriented text comparison are most obvious in the case of highly conventionalized text types. The conventions may concern all linguistic levels in the textual realization. At the syntactic level, for example, a parallel text analysis of instruction manu als shows that whenever the instructions are arranged in a sequence to indicate the individual steps of the action, imperatives are preferred in English texts, whereas in German texts infinitives are used (cf. Kußmaul 1990). Illustrative examples from a German and English instruction leaf let for hairdryers are “Check the position of the voltage selector switch before use” and “Darauf achten, daß die eingestellte Spannung mit der Netzspannung übereinstimmt”. Highly conventionalized lexical formulations are commonly found in contracts and treaties. Especially in the preamble, there are standardized phrases in each language which translators are expected to be aware of. Some of these phrases are given in figure 2: Figure 2. Standardized phrases in treaty preambles
English conscious
French conscients
desiring désirant recognizing reconnaissant
Spanish conscientes con el deseo reconociendo
German in dem Bewußtsein; eingedenk in dem Wunsch in Anerkennung; in (der) Erkenntnis
Before general recommendations for translation can be given, a large corpus of parallel texts of a specific text type must be analyzed in detail in order to discover regular, typical conventions and/or to set up proto types of text types (cf. Neubert 1985; Kußmaul 1995). In fact, finding equivalents in parallel texts is also a good exercise for translator training, something we practise at the University of Aston, in an Advanced Translation course for final year students. The students are asked to find exemplars of the text type in question, both in the source and in the target language, and then look for systematic regularities for the text type in both languages and cultures. Based on such comparisons they come up with their TT versions. Corresponding structures available for use in the TT are thus taken from actual usages in parallel texts. For example, students translated an instruction manual for a radio from German into English. A parallel text analysis had confirmed the conven tions about English imperatives and German infinitives mentioned above. Therefore, the students translated “Gerät einschalten und Lautstärke
86
Unity in Diversity
einstellen” as “Switch on the radio and adjust volume”. But something else became obvious from the parallel text analysis: the German text in cluded a piece of information reminding the readers that they have to get a licence before they can use the radio, and giving them the specific num ber they would need. Since this information is specific to the German source culture, there is no need to translate it into English. In contrast, English instruction manuals for electrical appliances include information about how to put on a plug (this has changed in the meantime, due to European Union regulations, which is evidence of the fact that text type conventions are historically determined and as such, apt to change). This culture-specific information unit need not be translated into German. Such findings point to differences in the macro- and superstructures of text types. However, analyzing parallel texts with such a focus, e.g. in formation arrangement, recurrent structures at different hierarchical levels, and deriving consequences for translation have only recently emerged (cf. Göpferich 1995). For another exercise, students translated a tourist brochure from Ger man into English. Looking at English parallel texts resulted in a closing formula “We will do everything possible to make your stay here a very pleasant one” for “Wir bemühen uns, Ihnen Ihren Aufenthalt so an genehm wie möglich zu machen”. Or “Alle zwanzig Meter eine neue, romantische Ecke ...” was rendered as “Every second minute you find a new, romantic corner ...” because in English tourist brochures, distance seems to be more frequently expressed in time, rather than in space. Focusing on translation as TT production also means that translationoriented text analysis should be understood as a TT-oriented analysis, i.e. as a prospective activity. It might therefore be useful to change the schema above because it gives the impression of translation as a linear process, moving from a ST to a TT, thus automatically giving a high status to the ST. This is also the main argument used by Holz-Mänttäri (1984) to speak of translatorial action (‘translatorisches Handeln’) instead of ‘transla tion’. She argues that the verb ‘translate’ requires a grammatical object, a reference to the ‘what’ that is (to be) translated, thus orienting the atten tion in a retrospective way. A retrospective orientation is also reflected in the bottom-up process of translation, i.e. working from source-language elements and transferring the text sentence by sentence, or phrase by phrase. A prospective view of translation, on the other hand, is related to a top-down process. It starts on the pragmatic level by deciding on the intended function of the translation and asking for specific text-typologi cal conventions, and for addressees’ background knowledge and their communicative needs. It puts the TT in the centre and makes it clear that the ST is but one of the factors influencing the make-up of the TT. This is illustrated in figure 3. Not all slots are filled in to symbolize that several
Schaffner: Parallel Texts in Translation
87
factors influence the TT production, and also that the weight of these factors can differ in each specific situation. Target communicative situation
Quasi parallel texts and new text types In the case of a highly conventionalized text type we can say that a TT that conforms to the target culture expectations will in all probability fulfil its intended function in an appropriate way (cf. ‘instrumental trans lation’ in Nord 1993). In translation practice, however, there are also texts for which there is not necessarily a corresponding text type in the target culture. This is proof of the fact that text types are dependent on the social situation of a culture. Elsewhere I have discussed the specific case of an information leaflet about how to make an emergency telephone call in Great Britain for which no corresponding German text type exists. To translate this text, however, information was found in similar text types (cf. Schaffner 1995). The last problem I wish to mention shows that internationalization processes in communication result in the phenomenon that TTs are pro duced by way of translation without the existence of a proper ST. The examples given here are multilingual but equally authoritative legal or po litical texts that have been inter-culturally and inter-lingually created, in inter- or supra-national organizations (such as the European Union, EU). One such document I looked at in more detail is the Manifesto for the Elections to the European Parliament of June 1994, adopted by the Con gress of the Party of European Socialists (PES) on 6 November 1993 (cf. Schaffner 1997). The document was produced by way of multilingual negotiation. Each party had put forward ideas on the basis of which a draft document was produced in German. During a lengthy negotiation process in the working languages English, French, German and Spanish, this draft was revised and amended. The final text was subsequently trans lated into the other languages of the EU member states. Since each of the four working languages had a role to play in setting up the final version,
88
Unity in Diversity
none of them can legitimately be called a source language. Equally, there was not one particular version of the copies in the working languages that served as a ST for the other language texts. A comparative analysis of the English, German, French, Italian, Span ish, Portuguese and Danish versions of the PES Manifesto revealed some features which are a reflection of specific conventions in the languages and cultures involved. Text-typological conventions are obvious, for ex ample, in the way the information is structured or framed. For the English version, lists are fairly typical, introduced by a phrase that usually ex presses some intention. The German text, however, prefers a structure where the list is framed, or bracketed, by a phrase expressing an inten tion, or a vision, cf.: This is the course for Europe we want to pursue: • create jobs, ...; • work for equality for women and men; • work for more democracy. Diesen Weg wollen wir weitergehen: Arbeit schaffen, ...! Frauen und Männer gleichstellen! Mehr Demokratie wagen! Das sind unsere Ziele. Of all the seven language versions, only the Spanish text follows the English pattern in this case. Another structural difference is the repetition of the introductory phrase: Thanks to us; • the structural funds have been doubled, ... • Higher standards of employment have been achieved across Europe, ... • nur durch uns sind die Mittel für die Strukturfonds verdoppelt... • nur durch uns sind die Standards im Bereich des Arbeitsschutzes europaweit deutlich angehoben ... • nur durch uns ... The French, Portuguese and Danish texts are similar to the German structure, the Italian one is similar to the English structure, whereas in the Spanish structure there is no phrase corresponding to ‘thanks to us’. In order to prove that such structural differences indeed reflect different text-typological conventions, one would have to analyze a number of parallel texts in the languages concerned. And here we encounter another problem: multilingually negotiated texts are a new text type. They are
Schaffner: Parallel Texts in Translation
89
multilingual and multicultural by their very nature, i.e. they are produced in a multicultural setting to fulfil a function in several countries. For such a text type, the concept of a parallel text does not seem to apply. One can, however, look at a similar text type in a particular culture to check whether specific text-typological conventions can be discovered. In this particu lar case, I looked at some election manifestos that had been produced by German and English parties for their local audience. Simple lists were found to be typical for the English texts; and also in the German texts, examples for framed lists were found. Simple lists, however, were used more frequently in the German texts than such framing structures. On the other hand, framing structures were not found in English texts. But it is too early to postulate this as a general characteristic feature (due to the very small corpus). Do these texts fit into a framework that defines translation as sourcetext-induced target-text-production? Concerning the original text production of the PES Manifesto, we have a specific case of mediated communication. There is no primary communicative situation in which a ST has fulfilled its communicative function for its source language ad dressees and where subsequently one TT has fulfilled its function for target language addressees in a secondary communicative situation. In fact, there is no ‘proper’ ST that was exclusively produced in one lan guage and culture, conforming to source culture specific text-typological conventions. For translation studies this means that the status of both the ST and the source language community as well as such notions as paral lel text, text-typological conventions, accuracy or appropriateness need to be reconsidered.
Conclusion A functional approach demonstrates that in the process of translation a target text may result that is neither completely identical with the con ventions of the source text nor with the conventions of (quasi) parallel texts in the target culture. This illustrates the complexity of the concept of the parallel text as well as the fact that there is a whole range of transi tional stages and fuzzy boundaries in defining text types in a source culture and in a target culture. In the daily practice of translation (as interlingual and intercultural communication), there are situations where a simplified view of ‘saying again in a target language text what had been said in a source language text’, or ‘getting meanings across’ does not apply. Such a retrospective view takes the ST as the yardstick for the quality of the TT, which results in a tendency to concentrate on linguistic aspects, i.e. micro-level deci sions. A prospective view, on the other hand, puts the TT into focus and
90
Unity in Diversity
concentrates on the appropriate functioning of the TT in a specific situa tion in a target culture. The quality of the TT then becomes a functional and dynamic concept. Functional approaches to translation, such as Skopos theory (cf. Reiß and Vermeer 1991), or the theory of translatorial action (Holz-Mänttäri 1984), offer a framework for such as shift from a retro spective to a prospective view of translation.
References Gläser, R. (1991) Fachtextsorten des Englischen, Tübingen: Gunter Narr. Göpferich, S. (1995) Textsorten in Naturwissenschaft und Technik. Pragmatische Typologie Kontrastierung Translation. Tübingen: Gunter Narr Verlag. Hartmann, R. R. K. (1980) Contrastive Textology, Heidelberg: Groos. Holz-Mänttäri, J. (1984) Translatorisches Handeln. Theorie und Methode, Helsinki: Suomalainen Tiedeakatemia. Kussmaul, P. (1990) ‘Instruktionen in deutschen und englischen Bedienung sanleitungen’, in R. Amtz and G. Thome (eds) Übersetzungswissenschaft. Ergebnisse und Perspektiven, Tübingen: Gunter Narr, 369-379. Kussmaul, P. (1995) Training the Translator, Amsterdam/Philadelphia: John Benjamins. Lenk, H. E. H. (1993) Praktische Textsortenlehre. Ein Lehr- und Handbuch der professionellen Textgestaltung, Helsinki: Universitätsverlag. Neubert, A. (1985) Text and Translation (Übersetzungswissenschaftliche Beiträge 8), Leipzig: Enzyklopädie. Neubert, A. and G. M. Shreve (1992) Translation as Text, Kent & London: Kent State University Press. Nord, C. (1993) Einführung in das funktionale Übersetzen, Tübingen: UTB. Reiß, K. and H. J. Vermeer (1991) Grundlegung einer allgemeinen Translationstheorie (= Linguistische Arbeiten 147), Tübingen: Niemeyer, 2nd edition. Schäffner, C. (1995) ‘Textsorten in der Übersetzung Analyse eines Übersetzungsbeispiels’, in I.-A. Busch-Lauer, S. Fiedler and M. Rüge (eds) Texte als Gegenstand linguistischer Forschung und Vermittlung. Festschrift für Rosemarie Gläser (LFS = Leipziger Fachsprachenstudien, Band 10), Frankfurt: Peter Lang, 47-56. Schäffner, C. (1997) ‘Where is the Source Text?’, in M. Schmidt and G. Wotjak (eds) Modelle der Translation. Festschrift zum 65 Geburtstag von Albrecht Neubert, Tübingen: Gunter Narr. Snell-Homby, M. (1988) Translation Studies: An Integrated Approach, Am sterdam/Philadelphia: John Benjamins.
Bilingual Reference Corpora for Translators and Translation Studies CAROL PETERS Istituto di Elaborazione della Informazione Consiglio Nazionale delle Ricerche, Pisa, Italy EUGENIO PICCHI Istituto di Linguistica Computazionale Consiglio Nazionale delle Ricerche, Pisa, Italy This paper discusses the potential of a bilingual reference cor pus as a translation resource. Such a corpus consists of sets of texts from pairs of languages that treat a given domain and can be contrasted because of their common features. This type of re source is useful because (a) it can provide translation equivalents and (b) it can allow for the identification of natural language lexical equivalents. The authors present procedures that have been developed at the Istituto di Linguistica Computazionale, Pisa, to construct and query bilingual reference corpora and to extract significant data for translation purposes and contrastive textual studies. The discussion is supported by examples of the different types of results that can be obtained.
Introduction It has long been recognized that the bilingual dictionary is an insufficient tool for translators and totally inadequate as a source of real world data for the purposes of translation studies. It has two serious drawbacks. One is that reported by Hartmann (1994:292): the semantic abstraction that is built into the lexical inventory of the dictionary has deprived each of these words of their natural con text, and the translator must compensate for the lack of contextual information from his/her own bilingual discourse competence, par ticularly in that most intractable area of ‘culture-specific’ vocabulary. Another, closely related problem is the division of each entry into a set of discrete senses, listing possible translations for each sense. As sug gested by Wilks (1977:204), real world practice is not so neat: Any careful thought shows that the actual usages of words in texts are difficult to assign in practice to one and only one set of pre sorted senses: it is not just that a usage may cover more than one, but it may be essentially vague between several.
92
Unity in Diversity
Unfortunately, when searching for the most appropriate target lan guage (TL) term for a given source language (SL) expression and attempting to insert it into a natural and meaningful syntactic-lexical con text, the ‘competence’ of the individual cited by Hartmann is not always able to make up for the lack of extensive contextual and usage informa tion in the bilingual dictionary. Another reference source is needed. Ideally the translator should have access to large systematic collections of reli able data attesting the usage of semantically equivalent lexical items across languages, according to variations in style, genre and domain, to supple ment the dictionary information. Such data collections would also be valuable for in-depth, exhaustive studies of the translation process. For this reason researchers are beginning to study the possibilities of a new reference tool: the bilingual text corpus. Whereas the bilingual diction ary supplies translation equivalents for lexical items considered in isolation as headwords (with a limited number of associated examples - generally of special cases), the bilingual corpus provides real world documented evidence of how utterances or samples of text in one language can be rendered in another. Bilingual corpora are thus valuable repositories of data on cross-language usage and can be exploited by a number of practi cal and theoretical applications. There are two very different types of bilingual corpora: parallel and comparable corpora. The former are sets of translationally equivalent texts, i.e. source texts and their translations. The latter consist of homo geneous sets of texts from pairs of languages which can be contrasted and compared because of their common features. They have been de scribed as collections of “texts which, though composed independently in the respective language communities, have the same communicative function” (Laffling 1991:81). (For a discussion of ‘monolingual compa rable corpora’, see Laviosa, this volume.) Thus, the texts in a bilingual comparable corpus are not translationally equivalent but are supposed to share certain basic features, such as period of time, topic, functionality, register, domain, etc. In particular, however, comparability implies the existence of a common vocabulary: a restricted sublanguage. Therefore, the two types of corpora provide different kinds of contrastive data: par allel corpora provide data on translation equivalents, whereas comparable corpora give information on natural language lexical equivalents within a given domain. Our claim is that they can both supply valuable data for the translator and for translation studies. In this article, we briefly describe the Bilingual Corpus System that has been implemented in an experimental version in Pisa. It includes pro cedures for managing and querying both parallel and comparable corpora. The system, together with a system for monolingual corpus management and interrogation, is one of the two main components of a prototype
Peters & Picchi: Billingual Reference Corpora
93
Translator’s Workstation described elsewhere (Peters and Picchi 1993). The other component is an integrated mono- and bilingual lexical data base system.
The parallel corpus procedures So far most of the systems studied that manage bilingual and multilin gual corpora work on parallel text and use statistically-based procedures to align the texts. The results of a query are normally given as parallel sentences. Church and Gale (1991) present a system of this type and also describe a word-based concordance tool in which the possible transla tions for a given word are discovered from the corpus on the basis of a pre-computed index indicating which words in one language corres pond to which words in the other. Other papers have suggested different ways of isolating and identifying translation equivalents with (Brown et al 1993) and without (Fung and Church 1994) pre-alignment. Many experiments have concentrated on the huge Hansards corpus of French/ English Canadian parliamentary debates and attention has tended to be focused on the computational processing of the bilingual texts rather than the retrieval of translationally interesting data. However, although such approaches provide good ways to prepare vast quantities of aligned translated texts for further machine processing, e.g. for NLP or MT ap plications, they do not really satisfy the needs of the typical human user. For example, both professional and trainee translators require easyto-use query procedures that allow them to access and retrieve detailed and precise information on L1/L2 translation equivalents. They want the data to be presented clearly, and to be able to filter out uninteresting information. In particular, they want to be able to find ‘new’ or ‘un known’ data on possible translation equivalents, i.e. information not already given in a bilingual dictionary, and to retrieve it in well-matched parallel contexts. Above all, they need procedures that are capable of pro cessing ‘difficult’ data efficiently, e.g. literary or journalistic material, in order to give interesting and exploitable results. We thus developed our system for parallel corpus processing with these requirements in mind: the system is easy for the non-computer expert to access and query, presents the results in a highly user friendly manner, and includes proce dures that allow users to construct parallel archives from their own data. For this reason, we have adopted a different approach to those men tioned above: the system is lexically rather than statistically based. In a first stage, we adopt a previously developed, electronic bilingual diction ary, and morphological components to link pairs of English/Italian texts on the basis of L1/L2 translation equivalents. This procedure is execut ed whenever a new pair of bilingual texts is added to the archives. Full
94
Unity in Diversity
details on how the text linking algorithm operates, how the ‘search zone’ (area of L2 corpus searched for L2 equivalents of a given LI word) is calculated, how false links are eliminated, and under what conditions fail ure is signalled, are given in Marinai et al. (1994) and will not be repeated here. It is sufficient to note that, from the user’s point of view, this opera tion is simple, rapid and, once a few preliminary instructions have been given, automatic. The L1/L2 links are then used by the bilingual text query system in a second stage to construct parallel contexts for any form or co-occurrence of forms searched in either of the two sets of texts. For each word or combination of words in one language (LI) searched by the user, parallel concordances from the sets of texts in both languages are constructed in real time and displayed on the screen. The LI context is constructed with the search word at the centre; this word is highlighted and, during the creation of its context, any translation links to the L2 text associated with the other words in the context are read. If the search word itself has an associated link, this is used to identify directly the corresponding word in the L2 text, which will also be highlighted and used as the central point for the construction of the L2 context; other words that have been linked in the paired contexts can be optionally displayed in a different colour. When there is no directly linked L2 form for the LI word being searched, then all the links for words found in the source context are used to calculate an ‘average’ value which identifies the central point around which the relative TL context will be constructed. The calculation of this ‘average’ value allows for the possibility of uneven concentrations of matched words in the contexts. The two linked forms which are closest to the point calculated as the middle of the target text context are displayed in a different colour, as indicators of the likely position of the translation equivalent. ‘Incorrect’ links between potentially falsely recognized trans lation equivalents which distort context calculation are identified and eliminated by the query system. The user can either search for single word forms or, using the morphological generator, for all the forms of a given lemma. Bilingual concordances of interest can be printed out or saved in a separate file for future reference. Figure 1 gives an example of the results of a search for parallel con cordances of any occurrence of the Italian lemma libertà in an English/ Italian parallel corpus. The search word is displayed between aster isks; where the procedures find a direct translation (i.e. a translation already contained in the bilingual dictionary), this is also displayed between asterisks, as in Concordance 3. The more interesting results are those in which no direct translation is found; in these cases, the L2 equivalent is to be found between the words marked in bold (see 1 and 2).
Peters & Picchi: Billingual Reference Corpora
95
Figure 1: Parallel concordances fo r the lemma ‘libertà ’
DBT - (Parallel Texts)
Search libertà
J. Joyce - The Dubliners
1) {1} “Tutti i ragazzi” disse “ce 1’ hanno l’innamorata.” Rimasi colpito dalla strana *libertà* con cui un vecchio come lui parlava di certe faccende. Pensavo fra me che quanto aveva detto dei ragazzi e ragazzine I-Dublin2.245 -1) {E}. “Every boy, he said, has a little sweetheart.” His attitude on this point struck me as strangely liberal in a man of his age. In my heart I thought that what he said about boys and sweethearts was reasonable. E-Dublin2.263 2) {I } permesso di andarsene non appena le donne avessero finito di prendere il té, e Maria attendeva con ansia quella sua serata di *libertà*. La cucina era tutto un luccichio: a detta della cuoca ci si sarebbe potuti specchiare sugli enormi paioli di rame. I-Dublinl0.5 -2) {E} given her leave to go out as soon as the women’s tea was over and Maria looked for ward to her evening out. The kitchen was spick and span: the cook said you could see yourself in the big copper boilers. E-Dublinl0.5 3) {1} fama levava, il verde vessillo spiegando, i guerrieri e i bardi incitava a levarsi agli occhi del mondo. *Libertà* soprattutto cercava, inseguendone 1* incantamento, e nel mentre afferrarla tentava, gli fu tolta da vii tradimento. Infamia I-Dublinl2.551 -3) {E} warriors raised Before the nations of the World. He dreamed (alas, ’twas but a dream!) Of *Liberty*: but as he strove To clutch that idol, treachery Sundered him from the thing he loved. E-Dublin 12.622
Instead of viewing the parallel concordances directly, users can choose to view a^preliminary analysis of the results. For any LI word queried, the L2 equivalents found in the parallel texts will be listed in descending order of frequency of occurrence for those words where direct links have been made. This list is followed by the number of parallel contexts in which no direct bilingual dictionary link has been found for the search word in the L2 text. Users can then select the particular contexts they wish to view without necessarily having to scan through them all. For example, they can choose to view only those concordances which give ‘new’ information, i.e. the L2 equivalent was not found in the bilingual dictionary. At the moment, the system runs on a sample set of Italian/English texts, chosen to cover a number of different language varieties, and in cluding extracts from scientific articles, university and school text books, magazine articles, short stories, novels and poetry.
The comparable corpus procedures A major criticism that has been made with respect to the results of analy ses based on translationally equivalent texts is that target texts are not true examples of natural language: a translation is always influenced by the source text. This means that while parallel text corpora are important
96
Unity in Diversity
sources for studying the translation process, the target language or L2 text can never be considered in itself as a true representation of that language in use. As stated by Hartmann (in press), “the translated text(s) cannot by definition share the full range of linguistic features of genuine texts produced in the respective target language”. For this reason, research ers have begun to use bilingual and multilingual comparable corpora as potentially more reliable sources for certain types of studies, e.g. studies of how a particular concept is rendered independently in different lan guages (the starting point being the idea and not an already existing text in a given language). We have thus decided to extend the scope of our bilingual corpus system by including a set of procedures for the analysis and extraction of significant data from comparable corpora. Parallel and comparable sys tems serve different purposes: with our parallel system users can retrieve examples of specific instances of how a given word or expression has been translated in another language, depending on context, argument, stylistic considerations, etc.; using the comparable system, they will be able to look for natural language examples of L2 lexical equivalents of a given word or expression in LI, independently of any translation link. By definition comparable corpora involve special domains or sublanguages; they are thus of particular interest to terminologists and technical, rather than literary, translators. Our procedures operate on sets of comparable texts in two different languages. We are currently working on Italian and English texts, and so far all work has been focused on nouns. The approach is based on the assumption that (i) words acquire sense from their context, and (ii) words used in a similar way throughout a sublanguage or special domain corpus will be semantically similar. It follows that, if it is possible to establish equivalences between several items contained in two different contexts, there is a high probability that the two contexts themselves are to some extent similar. It is important to stress that our aim is not to retrieve pre cise equivalences in L2 of the LI term under examination, but to isolate the set of contexts in the L2 corpus that has the highest probability of providing L2 correspondences to the LI input. Given a particular term or set of terms found in the texts in one language (LI), we attempt to iden tify contexts which treat the same topic in the texts of the second language (L2). To do this, we isolate the vocabulary related to that term in the LI corpus - hypothesizing that the word will be surrounded by a similar vocabulary in L2. A term, T, is thus selected in one set of texts, called LI (either set can be chosen as LI). For each occurrence of T in the LI set of texts, the system constructs a context window containing T plus up to n lexically significant words appearing to the right and left of T, but within the same
Peters & Picchi: Billingual Reference Corpora
97
sentence: strong punctuation marks (full stops and semi-colons) act as boundaries in the construction of these contexts. The value for n is set by the user. Words contained in a stop list are not counted. This stop list includes function words such as articles, pronouns, and prepositions, and also very frequent insignificant words which would create noise. It can be modified by the user so that certain frequent terms specific to the par ticular domain can be eliminated if necessary to improve performance. Morphological procedures identify the source lemma of each form co-occurring with the keyword T in the context windows. The set of signi ficant lemmas found in the context windows for T make up the vocabulary, V }, that characterises T in the particular LI corpus. The frequencies of the elements of W xare then computed and each is assigned a mutual infor mation (MI) value, which measures the significance of the association between the V 1 item and T (see Church and Hanks 1990). Using the MI index as an ordering element, we list Y i in order of decreasing signifi cance and set a threshold below which elements of V 1are not considered relevant and so can be ignored. Figure 2 shows the significant collocates for the Italian lemma libertà found in a set of Italian parliamentary de bates. There were 198 occurrences of libertà. For each collocate, the first column shows the MI value, and the second the number of times the col locate was found in the context windows for libertà when n - 5. Each collocate is given a gloss in English for the reader’s convenience. Figure 2: Significant collocates o f ‘ lib ertà ’
libertà 0000000 500.000 11.259 10.094 9.358 8.965 8.722 8.619 8.573 8.550 8.204 7.696 6.672 6.155 6.040 5.975 5.746
-
198 occurrences
-
16 significant cc
16 198ILIBERTA’ (freedom) 4ICONDIZIONARE (condition) 30IFOND AMENT ALE (fundamental) 5IESPRESSIONE (expression) 5ISINDACARE (inspect/control) 3IEFFETTIVO (effective) 4IPIENAIPIENARE (fulllto fill) 4IINDIVIDUARE (to identify) 4IBENEFICIARE (to benefit by/from) 3ILIMIT AZIONE (limitation) 3IGARANZIA (guarantee) 5IPRINCIPIOIPRINCIPIARE (principlelto begin) 4ICITTADINO (citizen) 5IRISPETTOIRISPETTARE (respectlto respect) 6IPOLITICA (politics/policy) 5ITRATTAREITRATTATO (to treatltreaty)
Next, using morphological analyzers and generators, and a bilingual lexical database, we construct an equivalent vocabulary V2 in L2 : for
98
Unity in Diversity
each element of V l we create a set of L2 translation equivalents, called L2 translation blocks. Each block contains the set of translations, retrieved from the bilingual lexical database, for one element of V l9 together with all possible forms of each translation. For example the L2 translation block for the Italian lemma garanzia includes the English forms guaran tee, guarantees, security, securities, surety, sureties. To each translation block, we assign a value equal to the MI value of the element repre sented by this translation block. These values are used to assign weights to the translation blocks to represent the probability of occurrence in the L2 texts of any of the members of that particular translation block when searching for expressions regarding our keyword, T. The procedure then searches the L2 corpus in order to identify words or expressions that can be considered as in some way lexically equivalent to our selected term T in the LI texts. This is done by searching for those contexts in L2 in which there is a significant presence of the L2 vocabu lary for T. The significance is determined on the basis of a statistical procedure; this procedure uses the number of V2 items found in the con text and the weights assigned to them, in order to assess the probability that any given L2 co-occurrence represents a lexically equivalent context for T, and to establish thresholds of acceptability. Although it is clear that the process of translating the LI vocabulary for T into L2 introduces a number of irrelevant terms, this does not normally affect the results as, if an L2 context is to be accepted as representative of a given LI term, it is necessary for a number of items from the L2 vocabulary for T to be present. The results are written to a file and listed in descending order of (i) the number of items from the LI context contained in each L2 translation block, and (ii) the sum of the MI values associated with these LI items. This file can be displayed on the screen and browsed, or printed out or saved for further consultation. Users can also access the entire text from which a selected context comes, simply by clicking on that context. Figure 3 gives an example of a query on our comparable corpus for the Italian lemma libertà. In order to save space we include here only the first 12 contexts, i.e. those calculated by the system as being most repre sentative of the use of this term in this particular corpus. As a test, we excluded the translations of the term given in the bilingual dictionary (liberty, freedom ) from the construction of the L2 contexts. The fact that a direct translation of libertà appears in a number of the results (Nos. 2, 3, 4, 5, 11 and 12) suggests that the system is performing well. Our system for querying bilingual comparable corpora is still in a pre liminary stage of development. Work remains to be done to refine the search criteria and increase the efficiency of the global algorithm. We feel, however, that our first results on real texts are encouraging and al ready demonstrate the validity of our approach. We are able to identify
99
Peters & Picchi: Billingual Reference Corpora
and retrieve contexts in a set of L2 texts which refer to a particular topic represented in LI by a given term, without recourse to an explicit transla tion equivalent for that LI term. Figure 3: Query results fo r ‘libertà ’ using an Italian/English comparable corpus
DBT - (Comparable Texts)
Search libertà
4 24.842 20 1) to make changes in funding *policy* in relation to *cities*, while still R e specting* the *principle* of =FE “FXAC93207ENC.0014.01.00”. 14 3 26.512 36 2) Respect for human rights and *fundamental* freedoms is *guaranteed* in the Member States by *effective* systems of =FE “FXAC93145ENC.0030.01.00”.32 3 24.462 38 3) as long as that legislation *guarantees* that the relevant *fundamental* *principles* and freedoms enshrined in the = F“FXAC93288ENC.0024.01.00”.27 3 23.830 38 4) as guardian of the Treaties. *Respect* for human rights and *fundamental* freedoms is *guaranteed* in the Member =FE “FXAC93145ENC.0030.01.00”.32 3 22.806 40 5) in the Treaty, such as the *principle* of non-discrimination, *respect* for the *fundamental* freedoms enshrined =FE “FXAC93040ENC.0013.01.00”.50 3 22.806 40 6) that all Community law ought to *respect* this *principle* because it is a *fundamental* right. Despite the =FE “FXAC93162ENC.0013.01.00”.40 3 22.806 40 7) countries must be based on such *principles* as *respect* for international law, human rights and *fundamental* =FE “FXAC93327ENC.0036.02.00”.26 3 22.741 41 8) in the drawing up of all *policies* is a *fundamental* *principle* of Com munity legislation. In view of this: 1. =FE “FXAC93065ENC.0034.01.00”.16 3 22.741 41 9) in Portugal. This discriminatory *policy* is in clear breach of the most *basic* *principles* and provisions of national =FE FXAC93095ENC.0014.03.00”.19 3 22.512 40 10) nationality which conflict with the *fundamental* *principles* of the EEC *Treaty*. It should be made clear that the =FE “FXAC93047ENC.0005.01.00”.27 3 22.512 40 11) under Article 115 of the EEC *Treaty* derogate from the EEC Treaty’s *basic* *principle* of freedom of movement of =FE “FXAC93264ENC.0031.01.00”.32 3 22.512 40 12) guarantees that the relevant *fundamental* *principles* and freedoms enshrined in the EEC *Treaty* will be observed. =“FXAC93288ENC.0024.01.00”.27
Final remarks The objective of this article has been to demonstrate that bilingual refer ence corpora can provide valuable sources of data for both practical translation work and theoretical studies. Translation of discourse is only possible if we know the equivalent structures in the target language. Ap propriate word selection requires far more than dictionary look-up; it must be possible to access information on actual text usage. Parallel and com parable corpora can provide the appropriate frame of reference, providing the user with selection pointers which function in a far more refined way than the general equivalence relations found in the bilingual dictionary. We have described procedures designed to provide user-friendly access to both types of corpora. The procedures have been tested on sample sets of Italian/English texts; however, they could be easily extended to cover other languages given the necessary lexical and morphological compo nents. The construction of high quality bilingual and multilingual corpora,
100
Unity in Diversity
sufficiently representative of the objects they aim at modelling (whether the entire lexicon or particular sub-sets of it) is clearly up to the user and dependent on the particular application. A number of such corpora al ready exist; co-ordination of efforts is recommended in order to ensure the reusability of data and avoid duplication of work. However, a discus sion on this topic is beyond the scope of the current article.
References Brown, P., S. Della Pietra, V. Della Pietra and R. Mercer (1993) The Math ematics of Statistical Machine Translation: Parameter Estimation’, Computational Linguistics 19(2):263-311. Church, K. W. and Gale, W. (1991) ‘Concordances for Parallel Text’, in Using Corpora , Proceedings of the 7th Annual Conference of the UW Centre for the New OED and Text Research, Oxford: OUP, 40-62. Church, K. W. and Hanks, P. (1990) ‘Word Association Norms, Mutual In formation and Lexicography’, Computational Linguistics 16(l):22-29. Fung, P. and Church, K. W. (1994) ‘K-vec: a New Approach for Aligning Parallel Texts’, in COLING 94 - The 15th International Conference on Computational Linguistics , Kyoto, Japan, 1096-1101. Hartmann, R. R. K. (1994) ‘The Use of Parallel Text Corpora in the Genera tion of Translation Equivalents for Bilingual Lexicography’, in Euralex 1994 Proceedings , Amsterdam, 291-297. Hartmann, R. R. K. (in press) ‘From Contrastive Textology to Parallel Text Corpora: Theory and Applications’, in R. Hickey and S. Puppel (eds) Language History and Language Modelling , Festschrift in Honour of Jacek Fisiak’s 60th Birthday, Berlin: W. de Gruyter. Laffling, J. (1991) Towards High-Precision Machine Translation - Based on Contrastive Textology, Distributed Language Translation 7, Berlin: Foris. Marinai, E., Peters, C. and Picchi, E. (1991) ‘Bilingual Reference Corpora: A System for Parallel Text Retrieval’, in Using Corpora, Proceedings of the 7th Annual Conference of the UW Centre for the New OED and Text Research, Oxford: OUP, 63-70. Peters, C. and Picchi, E. (1993) ‘Computational Tools for the Translator’, Netvaerk LSP SSP , Fagsprogsforskningen i Norden, Nyhedsbrev 7:1-10. Wilks, Y. (1977) ‘Good and Bad Arguments about Semantic Primitives’, in Communication and Cognition 10:181-221.
The English Comparable Corpus A Resource and a Methodology SARA LAVIOSA Department o f Language Engineering, UMIST, Manchester, UK A recent trend in translation studies is the use o f corpora fo r empiri cal and descriptive studies. The author discusses the design and compilation o f an English comparable corpus, which is a corpus made up o f two sets o f texts: one set that has been originally writ ten in English, and another which has been translated into English. This corpus is then used to study the phenomenon o f simplification, a feature hypothesized as being universal in translated texts. Spe cific features o f translational vs. non-translational texts that are analyzed and discussed include: range o f vocabulary, information load, and average sentence length.
The English Comparable Corpus (ECC) When referring to corpora, scholars in translation studies and corpus lin guistics use the term comparable in different ways. In my study I have adopted the definition proposed by Baker (1995), according to which a comparable corpus consists of two separate collections of texts in the same language. One collection contains texts originally produced in a given language, the other includes texts translated into that same lan guage from one or more source languages. To the best of my knowledge, the concept of comparable corpus has been applied in only two separate studies: Gellerstam’s (1986) computer-based analysis of Swedish translationese and Puurtinen’s (1995) manual analysis of the occurrence of finite and non-finite constructions in two non-computerized collections of translational and non-translational Finnish fairy tales and fantasy stories. My long-term objective is to create a representative, cohesive and re latively large English Comparable Corpus (ECC) for the systematic study of the linguistic nature of translation. In this article, I will concentrate on one aspect of my research; namely the investigation of simplification us ing one section of the ECC: the Newspapers Subcorpus. The com position o f the new spapers subcorpus
As shown in table 1, the ECC Newspapers Subcorpus contains articles from The Guardian and The European , grouped in four separate collec tions: two translational (TEC: Translational English Corpus) and two non-translational (NON-TEC: Non-Translational English Corpus).
102
Unity in Diversity
Table 1. Composition o f the ECC newspapers subcorpus
EEC N ew sp ap ers S u bcorpus The G uardian TEC
The G uardian N O N -TEC
The E uropean TEC P ressw atch
The E uropean N O N -TEC Pressw atch Sevenday
The Guardian collections The composition of The Guardian collections is shown in table 2. The authors and the translators are varied. The articles are complete transla tions of originals published in European national newspapers. The source languages (SLs) are varied. Table 2. Composition o f ‘The Guardian ’ collections The G uardian TEC
The G uardian N O N TEC
N um ber o f articles
102
102
Total w ord count
65,151
59,356
A verage w ord count per
638.73
581.92
Tim e span
19/5/94 - 15/12/94
19/5/94 - 15/12/94
N ew spaper section
G uardian E urope
Hom e N ews
Topic
P olitics
P olitics
article
The European collections The composition of The European collections is shown in table 3. The ar ticles in The European TEC consist of excerpts selected from the editorials of European national newspapers. The SLs are varied. The translators are two British female journalists who were aged 25 and 30 when the translations were done. The E uropean NON-TEC is made up of two subcollections. The Presswatch articles consist of excerpts from editorials of British, Irish and American daily newspapers. Most of them have been excerpted by
103
Laviosa: The English Comparable Corpus
the same two journalists who translate for Presswatch. The Sevenday texts are original, complete articles, which have been written by a variety of authors. Table 3. Composition o f ‘The European ’ collections The European NON-TEC
The European TEC Presswatch
Presswatch
Sevenday
Number of articles
64
64
64
Total word count
9,640
13,021
16,087
Average word count per article
150.62
203.45
251.36
Time span
27/5/93 30/11/95
27/5/93 2/11/95
5/10/95 23/11/95
Topic
Politics
Politics
Politics
The source languages
The Romance languages are the best represented in both TEC collec tions, followed by the Germanic group. There is a much greater variety of SLs in The Guardian than in The European. The space reserved for translations is also considerably smaller in The European , as shown by the total number of articles published during the selected time span and the average word count per article.
Simplification: a hypothesized universal feature of translated texts On the basis of contrastive analyses of translated texts (TTs) and their source texts (STs), a number of scholars have intuitively identified sim plification as a feature which is universal in all types of TTs, independent of the influence of the specific language pairs involved in the process of translation (Blum-Kulka & Levenston 1983; Vanderauwera 1985; Baker 1992). The aim of this study is to test three basic hypotheses which I consider to be consistent with simplification in translation. Statem ent o f hypotheses
The assumptions underlying my predictions are that a relatively limited range of vocabulary, a comparatively low information load and a relative ly low average sentence length are three different aspects of simplification
Unity in Diversity
104
in text production. As measures of vocabulary range and information load, I use an unlemmatized type/token/100 ratio and lexical density respectively. The former is the ratio of the number of different words to the number of running words in a text, calculated on chunks of 100 words of text at a time and then averaged out for the entire text. Lexical density, as defined by Stubbs (1986:33), is the ratio of the number of lexical words (i.e. running words minus function words) to the number of running words in a text. Both measures are expressed as percentages. H ypotheses o f sim plification:
•
•
•
The Guardian TEC has a lower type/token/100 ratio, lexical density and mean sentence length than The Guardian NON-TEC and these differences are independent of the SL variable. The European TEC Presswatch has a lower type/token/100 ratio, lexi cal density and mean sentence length than both The E uropean NON-TEC Presswatch and The European NON-TEC Sevenday and these differences are independent of the SL variable. The differences in type/token/100 ratio, lexical density and mean sen tence length between TEC Presswatch and NON-TEC Presswatch are lower than the differences between TEC Presswatch and NON-TEC Sevenday.
The last hypothesis is based on the assumption that comparing trans lational with non-translational texts that are both excerpts from original articles involves measuring the influence of one modification, namely the modification affecting the linguistic code. On the other hand, com paring translations of excerpts from STs with non-translational texts that are complete originals involves assessing the overall effect of two modi fications: one affecting the code, the other, the content.
Software for the automatic analysis of the ECC newspapers subcorpus To analyze the ECC newspapers subcorpus, three different pieces of soft ware were used: WordSmith Tools (Scott 1996), Microsoft Works for Windows™ Database and the Sharp EL-9200 Graphic Scientific Calcu lator. The concordancer from the WordSmith package was used to obtain the following statistics for each ECC text: (a) total number of running words; (b) type/token/100 ratio; (c) total number of function words; and 1In this chapter the terms ‘non-significant’ and ‘significant’ are used in their statistical, rather than their general meaning of ‘important’ or ‘considerable’. I am drawing atten tion to this distinction, because I would like to emphasize, in line with Woods et al (1986:127,130), that the significance level of the value of atest statistic indicates
Laviosa: The English Comparable Corpus
105
(d) mean sentence length. The database program was used to create two database files, one for TEC the other for NON-TEC. In each database record I included (a) de scriptive information about the individual text files; and (b) statistical lexical data. The Sharp Calculator was used to write and run two programs with which I carried out the parametric statistical tests for assessing the sig nificance of the various statistics.
Results concerning the hypotheses relating to The Guardian The difference in type/token/100 ratio between The Guardian TEC and The Guardian NON-TEC are found to be statistically non-significant1(see table 4). The differences in both lexical density and mean sentence length are found to be statistically highly significant. No statistically significant differences are found with regard to the three measures of simplification between three equal samples of 17 randomly selected articles translated from Germanic, Romance and Slavic SLs respectively. The hypothesis of simplification is confirmed with respect to lexical density and mean sentence length. A dditional results relating to The Guardian
Variance Variance is a statistical measure of the variability or dispersion of scores around the average value. The variance on all three measures of simplifi cation is lower for the translational texts (see table 4). It indicates the degree to which a group lacks homogeneity, so that the higher its value, the more heterogeneous the group is. The differences in the variance val ues recorded for type/token/100 ratio and lexical density are statistically non-significant while the difference in the variance recorded for mean sentence length is statistically significant.
Results concerning the hypotheses relating to The European TEC P ressw atch vs NON-TEC Pressw atch
The difference in type/token/100 ratio between TEC and NON-TEC no less and no more than the extent to which the research findings obtained are due to sampling variability. It is therefore recognized that the value of statistical hypoth esis testing rests solely on its ability to show how strong the evidence is for or against a given working hypothesis which is being tested in a single study. See LaviosaBraithwaite (1996) for details of the levels of significance obtained for my findings.
Unity in Diversity
106
Presswatch is statistically non-significant. The difference in lexical den sity is statistically significant and the difference in mean sentence length is statistically highly significant (see table 6). The hypothesis of simplifi cation is therefore supported on two measures: lexical density and mean sentence length. Table 4. The Guardian collections: scores on simplification and variance T he G uardian TEC
N O N -TEC
T Y P E /T O K E N /l 00 RATIO
7 3 .69254902
74.20794118
V ariance
8.455101346
10.721936
L E X IC A L D E N SITY
59.46273927
62.95240354
V ariance
8.527265098
10.099028
M EA N SE N T E N C E LEN G TH
20.96921569
23.6127451
V ariance
14.91228958
24.33237677
D ifference betw een TEC and N O N TEC -0.51 5 3 3 9 2 1 6
-3.48966427
-2.64352941
TEC Pressw atch vs N O N -TEC Sevenday
The difference in type/token/100 ratio between The E uropean TEC Presswatch and The European NON-TEC Sevenday is non-significant (see table 6). The differences in lexical density and mean sentence length are highly significant. Both discrepancies are greater than those found between TEC and NON-TEC Presswatch (see table 6). The hypothesis of simplification is therefore confirmed on two measures: lexical density and mean sentence length, but not on lexical variation. The last hypoth esis - which predicts a greater difference in simplification between TEC Presswatch and NON-TEC Sevenday, compared with the difference be tween TEC Presswatch and NON-TEC Presswatch - is confirmed for lexical density and mean sentence length, but not for type/token/100 ra tio (see table 6). Influence o f the SL on sim plification
No statistically significant differences are found either with respect to lexical density or type/token/100 ratio between three samples of 14 ran
Laviosa: The English Comparable Corpus
107
domly selected articles translated from German, French and Spanish/ Italian. A statistically significant difference associated with the SL vari able is found with respect to the mean sentence length. From the mean sentence length totals of the three groups it appears that the discrepancy is due to the scores of the mixed sample containing 6 translations from Italian and 8 from Spanish. These translations have, on average, longer sentences than those from German and French respectively. The hypoth esized independence of simplification in relation to the SL variable, is therefore confirmed for lexical density, but not for mean sentence length. A dditional results relating to The European
Variance TEC Presswatch has a statistically significant higher level of variance than NON-TEC Presswatch with regard to the type/token/100 ratio. The differences in variance for lexical density and mean sentence length are statistically non-significant. When the variance values are compared with those of NON-TEC Sevenday, they are found to be significantly lower for lexical density, significantly higher for mean sentence length and nonsignificantly lower for type/token/100 ratio. NON-TEC Presswatch has a significantly lower variance than NON-TEC Sevenday only with regard to the type/token/100 ratio (see table 5). NO N -TEC P ressw atch vs NON-TEC Sevenday
Interestingly, I find statistically significant differences on all three meas ures of simplification when I compare the non-translational excerpted articles with the comparable complete, original articles.
Additional results relating to The Guardian vs The European The Guardian NON-TEC vs The European N ON-TEC Sevenday
On all three measures of simplification The Guardian scores lower than The European. Statistically significant differences are found in lexical density and mean sentence length, but not in the type/token/100 ratio. These discrepancies are not as strong as the differences found within the individual newspapers between translational and non-translational texts on the one hand and translational excerpted texts versus original articles on the other. The Guardian TEC vs The European TEC
The difference in type/token/100 ratio is statistically non-significant. Both
Unity in Diversity
108
lexical density and mean sentence length are higher in The Guardian TEC and the differences are statistically significant. This finding seems to support the assumption expressed in the last hypothesis relating to The European Collections that when two modifications are taking place one affecting the linguistic code (translating) and one affecting the con tent (excerpting) - the impact on simplification is greater than when only one of either modification occurs. The importance of the result rests on the fact that it concerns two mediation processes taking place in two dif ferent newspapers. Table 5. ‘The European’ collections: scores on simplification and variance The European TEC
The European NON-TEC
Presswatch
Presswatch
TYPE/TOKEN/100 RATIO
73.93984375
73.33796875 74.53296875
Variance
20.065164
13.071863
LEXICAL DEN SITY
58.60578783
59.54017838 66.07182996
Variance
9.9886497
10.923595
14.051864
MEAN SENTENCE LENGTH
19.0334375
21.82265625
24.78453125
Variance
21.97446006
16.23789451
13.54239978
Differences between TEC and NON-TEC Presswatch
Sevenday
Differences between TEC Presswatch and NON-TEC Sevenday
+0.601875
-0.593125
-0.93439055
-7.46604213
-2.78921875
-5.75109375
20.744858
The Guardian TEC with The European NON-TEC Presswatch
There are no statistically significant differences in type/token/100 ratio or in lexical density. There is a significant difference in mean sentence length, but this is considerably smaller than the discrepancy found be tween the original articles of the two newspapers. These findings seem to suggest that with regard to lexical density and mean sentence length trans lating and excerpting are both associated with a relatively higher level of simplification independently of the newspaper type.
Discussion The Guardian collections
The TEC texts appear to exhibit a higher level of simplification than the comparable NON-TEC texts with respect to both the proportion of con
Laviosa: The English Comparable Corpus
109
tent words used and the average sentence length. On the other hand, lexi cal variation, as measured by an unlemmatized type/token/100 ratio, appears to be only negligibly lower. It can therefore be suggested that the translational process may not have a significant influence on the number of different unlemmatized words used, but it may affect lexical density and mean sentence length more noticeably. The lower variance of TEC articles with respect to mean sentence length seems to indicate that the translational texts are less idiosyncratic than the non-translational ones. I propose to call this phenomenon ‘con vergence’ in an attempt to convey the meaning of the clustering of a corpus of translations around the average value of a linguistic feature. Finally, the lack of evidence for differences in the simplification lev els associated with different families of SLs (Germanic, Romance and Slavic) seems to suggest that this feature may be a characteristic of trans lation p e r s e , regardless of the specific language pairs involved in the translating process. The European collections
TEC Presswatch vs NON-TEC Presswatch and NON-TEC Sevenday The evidence seems to suggest that the translational process may be as sociated with greater simplification with respect to the proportion of content words used and the length of the sentences, but not with regard to lexical variation, as measured by an unlemmatized type/token/100 ratio. This pattern is consistent with the one revealed by the analysis of The G uardian articles. Moreover, the co-occurrence of the modification of both the linguistic code and the content seems to have a greater impact on simplification than the sole occurrence of linguistic modification. The SL variable does not seem to have an effect on either the type/ token/100 ratio or the lexical density, but it does seem to influence mean sentence length. This result does not entirely conform to the evidence provided by the analysis of The Guardian articles, which shows no effect of the SL variable on any of the three measures of simplification. It must be pointed out that, given the very small size of the samples and the lim ited variety of languages represented, the results regarding both The Guardian and The European TEC texts could very well be due to the particular characteristics of the selected articles. Moreover, owing to the different distribution of the SLs in the two newspapers, it is not possible to create comparable samples and the results may very well be affected by these differences. It is therefore necessary to extend this type of inves tigation to a larger corpus and to a greater number of SLs in order to obtain more consistent patterns and consequently attempt to make firmer claims about the independence of the feature of simplification.
110
Unity in Diversity
NON-TEC Presswatch vs NON-TEC Sevenday The results concerning the isolated effect of excerpting suggest that this act of linguistic mediation is associated, like translating, with a relatively low proportion of content words and shorter sentences. Moreover, ex cerpting seems to be characterized also by a relatively narrow range of vocabulary, as measured by an unlemmatized type/token/100 ratio. Variance The modification of language and content together seems to be char acterized by a significantly higher level of homogeneity with regard to lexical density. This is consistent with the pattern revealed by The G uard ian TEC articles, which were less idiosyncratic on the same measure of simplification (although not significantly). This finding suggests even greater homogeneity when two language mediated events take place at the same time. On the other hand, the co-occurrence of the processes of translating and excerpting is associated with a higher level of idiosyncrasy with re gard to mean sentence length. One can only speculate very tentatively on the possible reasons for such a finding. It may be related to the influence of the SL or to the particular interaction of these two acts of language mediation. The replication of the same analysis on a much larger sample may in future help the researcher to suggest a plausible explanation for this result. The Guardian vs The European collections
The original articles of the two newspapers appear to be particularly different with respect to lexical density and mean sentence length, but these discrepancies are less marked than the ones found between their respective translational and non-translational texts. Moreover, the differ ences between the original articles of The Guardian and The European do not seem to disrupt the patterns of simplification exhibited by their translational articles. In fact, when the analysis cuts across the newspa pers’ boundaries, we still find that one modification - that of language appears to have less impact on lexical density and mean sentence length than the co-occurrence of language and content modification. This is con sistent with the finding that in the same newspaper ( The European) the modification of both the linguistic code and the content has a greater impact on simplification than the sole modification of content. More over, the translated articles of The Guardian and the excerpted articles of The European exhibit fairly similar levels of simplification, despite the
Laviosa: The English Comparable Corpus
111
fact that they belong to different newspapers. This, again, is consistent with the finding that in the same newspaper (The European) excerpting is associated with a simplification pattern similar to the one that character izes translation. The above evidence suggests that the two types of language mediation - translating and excerpting - exhibit similar pat terns of simplification and that these features are independent of the stylistic differences of the newspapers to which they belong.
Conclusions and suggestions for further research On the basis of the present study, I think I can assert with reasonable con fidence that the design and analysis of a comparable corpus can be very productive both in terms of providing insights into the translational fea ture of simplification and in terms of giving rise to new and more precise hypotheses about the elements that may affect this feature. Moreover, the availability within the newspapers subcorpus of different types of trans lational and non-translational articles has enabled me to explore the linguistic nature of translated text, not only in its traditionally assumed form of a complete target language text derived from an equally com plete source language text, but also as “selective text” (Sager 1994:178). Providing the corpus is enlarged so as to include texts translated from a greater variety of SLs, I suggest pursuing four further lines of enquiry: (a) the investigation of the influence of the SL variable on the level of simplification exhibited by the translated articles; (b) the calculation of the proportion of core or basic vocabulary in TEC and NON-TEC texts, using published lists of words that constitute the basic vocabulary of Eng lish (Stubbs 1996:98-115); (c) a detailed comparative analysis of the function words used by the ECC newspapers articles, with a view to ex ploring the possible links between simplification and explicitation in translation (Blum-Kulka 1986; Shlesinger 1995; Baker 1996; LaviosaBraithwaite 1998); and (d) a study of the influence on simplification of different sets of variables, such as those relating to the translator (e.g. gender, age, employment status, nationality and language direction), those relating to the extra-linguistic features of the STs and TTs (e.g. date of publication, publisher, country), and those concerning paralinguistic fea tures (e.g. diagrams and pictures).
Acknowledgments I would like to thank Dr. Mike Scott, University of Liverpool, for giving me permission to use a copy of WordSmith Tools for research purposes; Professor Michael Stubbs, University of Trier, for providing relevant lit erature; Mike Braithwaite for writing the programs used to carry out the
112
Unity in Diversity
statistical tests; and the Department of Language Engineering, UMIST, and the Sidney Perry Foundation for supporting my research.
References Baker, M. (1992) In Other Words, London: Routledge. Baker, M. (1995) ‘Corpora in Translation Studies: An Overview and some Suggestions for Future Research’, Target 7(2): Amsterdam/Philadelphia: John Benjamins, 223-243. Baker, M. (1996) ‘Corpus-Based Translation Studies: The Challenges that Lie Ahead’, in H. Somers (ed) Terminology, LSP and Translation: Stud ies in Language Engineering in Honour o f Juan C. Sager , Amsterdam/ Philadelphia: John Benjamins, 175-186. Blum-Kulka, S. (1986) ‘Shifts of Cohesion and Coherence in Translation’, in J. House and S. Blum-Kulka (eds) Interlingual and Intercultural Com m unication: D iscou rse and C ognition in Translation and Second Language Acquisition Studies , Tübingen: Gunter Narr, 17-35.
Gellerstam, M. (1986) ‘Translationese in Swedish novels translated from English’, in L.Wollin and H. Lindquist (eds) Translation Studies in Scan dinavia , Proceedings from the Scandinavian Symposium on Translation Theory (SSOTT) II Lund 14-15 June 1985 (Lund Studies in English 75), Lund: CWK Gleerup, 88-95. Laviosa-Braithwaite, S. (1996) The English Comparable Corpus (EEC): A Resource and a M ethodology fo r the Empirical Study o f Translation , PhD Thesis: Department of Language Engineering, University of Man chester Institute of Science and Technology (UMIST). Laviosa-Braithwaite, S. (1998) ‘Universals of Translation’, in M. Baker (ed) Routledge Encyclopedia o f Translation Studies , London: Routledge, 288-291. Puurtinen, T. (1995) Linguistic Acceptability in Translated Children's Lit erature (University of Joensuu Publications in the Humanities No 15), University of Joensuu. Sager, J. C. (1994) Language Engineering and Translation: Consequences o f Automation, Amsterdam/Philadelphia: John Benjamins. Scott, M. (1996) WordSmith Tools , Oxford: Oxford University Press. Shlesinger, M. (1995) ‘Shifts in Cohesion in Simultaneous Interpreting’, The Translator 1(2): 193-214. Stubbs, M. (1986) ‘Lexical Density: A Technique and Some Findings’, in M. Coulthard (ed) Talking about Text (Discourse Analysis Monograph No 13), University of Birmingham: English Language Research, 27-42. Stubbs, M. (1996) Text and Corpus Analysis: Computer-Assisted Studies o f Language and Culture , Oxford: Blackwell. Vanderauwera, R. (1985) Dutch Novels Translated into English: The Trans formation o f a “M inority” Literature , Amsterdam: Rodopi. Woods, A., P. Fletcher and A. Hughes (1986) Statistics in Language Stud ies , Cambridge: Cambridge University Press.
Section Four Computer-Aided Translation
Page Intentionally Left Blank
Practical Experience of Computer-Aided Translation Tools in the Software Localization Industry SHARON O’BRIEN International Translation and Publishing Ltd., Bray, Co. Wicklow, Ireland Software localization involves the translation o f complete software packages, including manuals, code and help systems. Newly re leased versions o f such packages often use previous versions as a basis fo r improvement. Obviously, software companies do not wish to pay fo r the same translation twice; they expect the localization companies to come up with ways o f leveraging the previously trans lated material into the new version o f their product. Until recently, the only way o f doing this was by comparing the documents and cutting and pasting. More recently, however, sophisticated Com puter Aided Translation (CAT) tools have come onto the market, which provide an alternative and more efficient way o f leveraging translation. This article discusses practical experience o f working with such tools in software localization, and points out some o f the advantages and challenges associated with them.
Introduction to translation in the localization industry Software localization involves translating and adapting software appli cations from one language, usually English, into other languages and cultures. A software application usually consists of three components: the software itself, which consists of code containing text strings that must be translated; the documentation (manuals); and the on-line help. The on-line help files are most often in Rich Text Format (RTF), and can be translated, for example, using Microsoft Word for Windows. The format of the documentation varies, but the most common format is a Word for Windows ‘doc’ file or a file that has been created in a desk-top publish ing package such as Framemaker or Ventura Publisher. The translation of a Word doc file is straightforward from a technical point of view. Most translators are familiar with this word-processing package or one similar to it. The translation of files created in desktop publishing packages, how ever, places more technical demands on the translator and necessitates an understanding of desk-top publishing packages. The translation of text strings in software files is probably the most difficult from a translator’s point of view because the text strings are often surrounded by software code which must never be altered or deleted. Translation of software re quires translators to familiarize themselves with a resource editor, which
Unity in Diversity
116
is an application used by programmers to edit code.
CAT tools in localization Software packages are constantly being improved and updated, and soft ware companies are increasingly aiming for simultaneous release of their products in many languages. The newly released version often uses the previous version as a basis for improvement. Therefore, some of the text, code and help system will be the same in the new version. Obviously, software companies do not wish to pay for the same translation twice; they expect localization companies to come up with ways of leveraging the previously translated material into the new version of their product. Until recently, the only way of doing this was by manually comparing the documents and cutting and pasting. More recently, however, sophistica ted Computer-Aided Translation (CAT) tools have come onto the market, which provide an alternative and more efficient way of leveraging trans lation. The choice of CAT tools currently available on the market is significant and growing, many boasting magical solutions to complicated questions. Finding the right tool to match one’s needs can often require much investment in terms of time and money, and even then there is no guarantee that one tool will meet all the demands put on it, especially by the localization industry. One type of CAT tool that meets some of the needs of the industry is the Translation Memory tool. Examples of trans lation memory tools include the Translator’s Workbench (for DOS or Windows) developed by the German company Trados, IBM’s Transla tion M anager (for OS/2 or Windows) and Globalware’s XL8. Most translation memory tools require a 486 PC (this is the minimum specifica tion - a Pentium chip provides better performance) with at least 8MB of RAM. Most tools run under Windows 3.1 or Windows 95, except XL8 (currently still DOS-based) and the OS/2 version of Translation Manager.
What is a translation memory? A translation memory (also called a ‘sentence memory’) consists of nu merous translation units. A translation unit is made up of a source sentence and its translated equivalent, for example: To make efficient use of these four buttons, organize your Library so that the first four views are the ones you use most frequently . Pour une utilisation efficace de ces boutons, organisez la bibliothèque afin que les quatre vues utilisées le plus fréquemment apparaissent en premier. The translation memory can be constructed either while a translator is translating a text or before or after the translation takes place. While a
O'Brien: CAT Tools in tOhe Localization Industry
117
translator is using a translation memory system, the translation unit is added to the translation memory, which acts like a database, storing the translation units. If a sentence has been translated once , it is then avail able in the translation memory and if that sentence occurs again in the text, the previous translation is suggested to the translator automatically. The translator has the choice of accepting the previous translation or edit ing it if the context requires change. The translation memory can propose perfect matches or fuzzy matches. A perfect match is identical to the sentence the translator is currently translating, both linguistically and from a formatting point of view. In the Translator’s Workbench for Windows, for example, a perfect match is indicated by the figure 100% which is displayed in the match window. A fuzzy match, on the other hand, is similar, but not identical, to the current sentence. A good translation memory system will highlight the differences between the sentences using colour codes and the percentage match is again indicated in the match window. A fuzzy match can be anywhere between 1 and 99%. From our experience, we have seen that even a 40% match can present useful translation information to the translator. A translation memory system also allows for batch translation. This means that you can run a set of files through a translation memory and the system will automatically replace any matches it finds with the trans lations stored in the translation memory. Material which is not translated in batch mode is translated by a translator and then the entire text is proof read and edited by the translator. If the translator makes changes to any matches that were inserted automatically, these changes can subsequently be implemented automatically in the translation memory, thereby keep ing the memory as up to date as the translation itself. If terminology needs to be changed throughout the text, and the same term occurs many times, it is possible, and often much quicker, to make the terminology change in the translation memory units by searching for the incorrect term and replacing it with a new term. The documents are then automatically updated with the changes which have been made in the translation memory.
Terminology capabilities in CAT tools The most favoured CAT tools on the market also have powerful termi nology capabilities. For example, XL8 has a glossary manager, Translation Manager has a dictionary list and the Workbench has a separate glossary manager called M ultiterm. In the localization industry, a client-approved glossary is one of the most important starting points for a localization project. Often, the source text (usually in English) glossary is created by extracting the terms and text strings from the software files. XL8, having been developed specifically for the localization industry, is commonly
118
Unity in Diversity
used for this task since it can protect the software code surrounding the text. Once the terms are translated and the client approves the termi nology, the glossary can be imported into the CAT tool’s terminology manager. While the translator is using the translation memory to lever age previous translation into the new text, the glossary recognizes terms in the sentence currently being translated and proposes the terms to the translator who can simply paste them into the translated sentence. While the terminology tools generally allow the user to input detailed informa tion such as gender, plural form, examples of usage, etc., client-approved glossaries in the localization industry often contain only the source term and the target term. In some circumstances, a reference to the source of the term might also be included, if, for example, the same term has two or more different translations depending on the context. Some of the rea sons for this type of glossary format include the following: (i) the required turn-around time in the localization industry is often so short that it does not allow for the preparation of detailed glossaries; (ii) the terminology used (even by the same client) can change rapidly, warranting new glos saries each time the client has a product localized; and (iii) the translator, who also has to produce very fast turn-around times, is interested only in the client-approved translated term and the context in which a term can occur if there is more than one translation for the same term.
Alignment As already mentioned, a translation memory can be created either dur ing, before or after translation. The process of creating a translation memory before or after translation is called alignment. Alignment is the automatic process of comparing a source file and the equivalent trans lated file, matching the sentences one by one and binding them together as translation units in a translation memory. An automatic alignment tool, such as Trados’ Talign can be used for this process. Since the task of aligning translated text automatically is a complicated one, there are some limitations, and the results depend on the suitability of the files for align ment. For example, the source text and target text must have a similar, if not identical structure. Some alignment programs can cope with single source sentences that have been translated as two target sentences or two source sentences that have been translated as one target sentence. They use certain clues, such as the number of characters in the sentence, acro nyms, numbers, formatting, etc., in order to guide the alignment process. In addition, special dictionaries containing abbreviations and keywords can be searched in order to aid the alignment process. Inevitably, some text is misaligned. There are ways of dealing with this too, however. Talign, for example, will put misaligned sentences into a separate file, so that they do not get included in the translation memory. Naturally some
O'Brien: CAT Tools in tOhe Localization Industry
119
misalignments are overlooked and the translation memory requires checking. This can be done either before translation begins or during translation where a misaligned unit is corrected by the translator. A trans lation memory is always more accurate when it has been created by interactive translation as opposed to automatic alignment, but alignment can produce a reasonably accurate translation memory which can be used as a start-up.
What are the advantages associated with CAT tools? CAT tools offer several advantages to the three main players in the lo calization industry: the client, the service provider (localization agency) and the translator. For the client
When clients update their products, they re-use code and text instead of re-creating what they have already produced. Localization companies are expected to do the same, i.e., to re-use existing translations. CAT tools offer a way of doing this by storing the translation in a translation memory. Our experience has shown that anything from 10% to even as high as 70% can be leveraged from translation memories. The figure for leveraging depends on how well the files are set up and on how much of the text has been changed. The client pays the normal rate for translation of the new text and a different rate for proof-reading and editing the old text. This means that the client’s costs can be cut. It should be pointed out that the advantages of long-term investment in CAT tools, where a large number of client-specific translation memories can be built up, far outweigh those of short-term investment. For example, creating translation memories by aligning previous translations might afford a client 40% leveraging from that previous translation. If the client continues to invest in the process, double that amount of leveraging may be achieved over time. As the translator does not have to re-translate what is already trans lated or to cut and paste from a previous translation (a tedious and time-consuming task), the through-put time for the translator is increas ed, allowing the client a faster time-to-market. The re-use of existing translations also allows for greater consistency between different versions of the same product and between different products developed by the same client. F or the localization company
Again, through the re-use of translation, a localization company can guar antee a reduction in turn-around time and an increase in consistency to the client. They can also translate a greater number of words per annum
120
Unity in Diversity
thanks to increased translation speeds. Our experience has shown that through-put figures can increase significantly when the circumstances are right. Altogether, CAT tools allow a localization company to offer an improved, more competitive service. For the translator
Traditionally, machine translation and even computer-aided translation were seen as a threat to the translator. However, the re-use of previous translation in the localization industry simply frees up more of the trans lator’s time allowing him/her to translate a greater number of words per annum in total. CAT tools allow for greater consistency in terminology and style. Also some CAT tools (like Trados’ Workbench) make it possible to share a translation memory on a network, allowing several translators access to the same translation memory at the same time. This ensures consistency throughout very large jobs consisting of several hundred thousand words. There is the added advantage of repetition matching where, if the same passage occurs in different manuals, and the translators are sharing a translation memory, it is possible that sections of text will have been trans lated by one translator for one manual, allowing another translator to simply leverage that translation into their document when they come ac ross the same passage of text. One major criticism of CAT/MT tools in the past was that the edit ing environment was not what the translator was used to and was, indeed, user-wnfriendly. Recent developments (e.g. Workbench) allow the translator to translate in Word, which is a familiar environment, while the translation memory system sits on top of, or even behind, the word processor. Unlike traditional MT systems, the translator feels very much in con trol of the translation procedure when using computer-aided translation tools as opposed to machine translation tools, since there is no linguistic parsing using grammatical rules.
What are the challenges posed by CAT tools? As there are advantages to be gained by the three main players in the localization industry, so too are there challenges to be faced. Challenges f o r the Client
CAT tools represent a relatively new technology. Clients who have not used this type of technology before often require some education as to the actual capabilities of CAT tools. Although reduced cost through re use of translation and faster turn-around times are real possibilities, initial
O'Brien: CAT Tools in tOhe Localization Industry
121
investment of time is required, so as to facilitate the creation of trans lation memories. Clients must be committed to CAT in the long-term if they are to see a return on their investment. The set-up of the client’s source files must be of a high quality in order to get the most efficient use from CAT tools. Failing this, some preparation work is necessary. The client must be willing to accept and implement recommendations for improvement to their file set-up. Similarly, if translation memories are to be built using alignment tools, the quality of the translated files must be acceptable to the transla tors who have to work with the resulting translation memories. The existence of several different tools, which often do not allow for easy transition from one tool to another, means that a client has to choose one CAT tool over another. Challenges fo r the localization com pany
Two of the greatest challenges for a localization company are resistance to change and glorification of the capabilities of CAT tools. These tools represent a change in the recognized way of doing things. Once the fear of this kind of change has been overcome and the real capabilities of CAT technology have been acknowledged, two of the greatest challenges have been faced. Significant investment is required from a localization company so that they can offer an efficient CAT service to all clients. Since the use of CAT tools affects not only translators, but also almost all other players in localization, the company must invest time and money in training for their translators, engineers, DTP specialists and project managers. Lo calization companies must also be ready to invest in research in order to customize the tools to their own needs and to make them more efficient. The introduction of CAT tools also calls for significant investment in terms of the equipment required to run the software (usually top of the range PCs and a stable network). If companies do not have the appropri ate equipment, the tools will not function properly and might even cause a drop in productivity. Finally, if companies are to derive maximum ef fectiveness from CAT tools, repeat business from clients is essential. Challenges fo r the translator
Again, the resistance to change can represent a major impasse when it comes to using translation tools. CAT tools must be presented to the trans lator as an a id , not a replacement for them or a hindrance to them. Proper training is imperative. This helps to overcome the fear of us ing a new tool and allows for more efficient use of the application(s). Since one tool can be more efficient at one task and another tool can be
122
Unity in Diversity
more efficient at another task, the translator must be familiar with more than one CAT tool. The editing environment for CAT tools has been criticized as being user-unfriendly in the past. However, this is changing and some can now be used in a familiar word-processing environment. Also, CAT tools have the reputation of being unstable and unreliable. However, proper use and constant improvement of the tools can prevent such problems. Developers of CAT software must be willing to listen to their users and ready to solve their problems rapidly.
CAT technology in the future CAT tools will undoubtedly touch the lives of most translators in the next couple of years. The translator most likely to benefit from their use is the one who works in a large translation agency or a large company employ ing in-house translators. It is in this type of scenario that the use of CAT tools is most efficient. As previously mentioned, translation memories can be shared on networks, where each translator can make use of trans lations produced by their colleagues, and consistency is guaranteed.
Translation Memories: Insights and Prospects MATTHIAS HEYN TRADOS Benelux S.A., Brussels, Belgium Translation memory systems are a type o f computer-aided transla tion tool that allow previous translations to be recycled in the course o f new translation jobs. The use o f this technology in the software localization industry has received some attention, but relatively lit tle is known about the growing body o f users outside the software sector. And while the basic priniciple o f translation memories is easily understood, state-of-art interfaces to such systems often be lie the complexity o f the technology beneath. This article aims to give an overview o f the many types o f user o f translation memories, and to link user profiles with different functional extensions o f the technology. Some o f the more technical aspects o f translation memo ries are then discussed. Finally, we flag a number o f issues that are beginning to emerge with the growing use of translation memories and that affect translators and technical writers alike.
Introduction In the past, automation of the professional translation process was usually associated with the use of machine translation (MT), but the situ ation has changed significantly in the last few years. Today, the keywords are computer-aided translation tools (CAT tools) and, especially, trans lation memories. O ’Brien (this volume) explains the basic concepts associated with CAT tools. Such tools, in most cases the integration of several functions in one workbench, are becoming standard in profes sional translation. CAT tools are now used in almost every type of translation work: political, administrative, technical, advertising and biographical, to name just a few. Whereas the general idea of a translation memory is fairly simple, the practical realization of a functioning product is rather complex. This has mainly to do with the subtasks that such a system has to perform. Trans lation memories involve many aspects of information science and linguistics, including database design, retrieval technology, mapping of complex data and text structures, client-server architecture, networking, support of language-dependent phenomena (character sets, tokenization, morphology, syntax), and software ergonomics. They represent an inter esting type of application, one which appears to users as a rather simple interface, but which has underneath a very complex internal functioning. Up to now, little attention has been paid to the needs of the various
124
Unity in Diversity
users of CAT tools. We can distinguish between a kernel set of functions in a translation memory and user-specific functional extensions. As the application area of translation memories continues to broaden, the func tional extensions of such systems become more and more varied. In what follows, we first identify the various factors that make the use of CAT tools attractive, and we describe different user profiles and needs. We then discuss technical aspects of CAT tools against this background. We conclude by addressing briefly a number of emerging non-technical is sues in the area of CAT tools, as well as the prospects for future development.
The benefits of using CAT tools Three benefits of CAT tools are usually cited by tool vendors (and see also O’Brien’s (this volume) section on ‘advantages for the client’): large quantities of texts can be translated faster (Quantity Argument); the qual ity of translation is increased (Quality Argument); and subsequent similar translation projects can benefit from earlier work (Re-usability Argument). In what follows, we elaborate on these rather general statements and de scribe in more detail some factors that play a major role in the application of CAT tools. These factors can then be used to distinguish between dif ferent user needs. By their very nature, translation memories find their main application in the translation of repetitive text material. It is important to distinguish between internal repetitions in a document itself and external repetitions where the repetitions are inherent to a family of documents, as happens with updated translations. We will call this the repetition factor. Translation memories enforce greater consistency in translation espe cially when they are integrated with a terminology database system. We will call this the consistency factor. Every translation unit can be ac companied by several types of information, for instance: creation user, creation time, update user, update time, subject code, notes, etc. This leads to an improvement in the quality of translation because revised and approved wordings are re-used. It is like using a translation from an authorized reference, and, as such, leads to standardization across translations. This factor will be called the reference factor. From a linguistic point of view, a translation memory can be described as a bilingual parallel corpus. In the case of systems that allow more than one source or target language, we can speak of multilingual parallel cor pora. Such corpora can be used to retrieve a translation unit, by searching for one or several keywords. This function is commonly referred to as concordancing, more precisely bilingual concordancing. Translation memories can be seen as a rich source of implicit terminology (in con trast to the explicit terminology stored in term banks). In this sense,
Heyn: Translation Memories - Insights & Prospects
125
translation memories can, and do, compete with term banks. This factor will be called the concordance factor. Terminology recognition, that is the automatic searching in an associ ated term bank for terminology in a source translation unit, plays a key role in CAT tools. Terminology recognition should not be confused with terminology extraction, which means the automatic extraction of termi nology from text material. Terminology recognition obviates the need for manual searches in databases; the system automatically draws the translator's attention to the relevant terms. CAT tool users thus benefit in two ways: they can keep track of specialized terms; and the retrieval of terminology can be manual or automatic. This factor will be called the
terminology factor. CAT tools can create resources automatically in three ways: firstly by creating a translation memory out of existing parallel texts in a process known as sentence alignment; secondly, by creating of a list of term can didates in one language to be introduced into the term bank system (monolingual terminology extraction); and thirdly, by creating a list of term-pair candidates from source and target texts to be introduced into a term bank system (word alignment or bilingual terminology extraction). We will refer to this factor as the resource creation factor.
Profile of CAT tool users As has already been mentioned, the general market for CAT tools is broad ening. In particular in countries where there is more than one national language or where translation costs are high, there is greater acceptance of software intended to rationalize the translation process. At the level of individual industries, it is well known that CAT tools have been used extensively in software localization (see O’Brien, this volume), but this is not the only sector where CAT tools find application: international marketing strategies and product liability laws that call for proper local ized documentation for targeted markets mean that such tools are now in use in areas as diverse as the pharmaceutical and aeronautics industries. While the repetition and consistency factors are of particular importance in localization and other industries, banks, insurance companies and le gal firms tend to place special emphasis on the terminology, reference and consistency factors. In military applications, there is the added re quirement of document confidentiality, which has implications for the way in which translation units can be stored. In the multimedia sector, the terminology and concordancing factors, as well as the ability to han dle HTML documents, are of utmost importance. Terminology is particularly important in the context of European and other international organizations, but traditional approaches to term bank creation and main tenance are complex and very costly. Here the concordance factor, in
126
Unity in Diversity
effect a by-product of the use of CAT tools, comes into play: during a test phase of Trados’s Translator's Workbench fo r W indows at the Euro pean Commission it turned out that concordance access to a translation memory with only 28,000 translation units was in many cases more help ful for retrieving terminology than access to Eurodicautom, the world’s largest multilingual term bank, with over 1,300,000 entries. In all inter national institutions, but especially those with standardized documents, the repetition factor plays an important role, but subphrase repetition, i.e. repetition within a translation unit, remains a problem for commer cial translation memory systems. Translation agencies that handle several clients are highly dependent on their clients when it comes to the format and type of document they have to translate. In some cases, current CAT tools do not yet support the format in question; in other cases, documents are not even available in machine readable form. Whereas in other areas the CAT tool can be ‘tuned’ to the text type in question, in translation agencies CAT tools must be more flexible. The management of CAT tools that this entails has led in the recent past to a new professional profile in translation agencies, that of the IT Manager. In the case of agencies that do not have in-house revisers, functions like pretranslation and off-line updating are required of a CAT tool, and users of term banks must have access to printing or electronic publishing facilities. Freelance translators, traditionally conservative when it comes to capital investment and thus less likely to use CAT tools, will find over the coming years that they will have to work with documents that have been pre-processed using these tools. They may even be temporarily forced into the use of a CAT tool by their work suppliers. Finally, two other user groups may begin to emerge in the near future: terminologists (primarily interested in the concordance and advanced re source creation factors), and non-professional users.
Technical issues in translation memories A translation memory is a database that stores translation units. This sim ple definition is complicated by a number of issues that are discussed below, starting with the thorny issue of similarity. Coping with sim ilarity
When are two sentences similar? This is a very tricky question. There could be misspellings, differences in the formatting, differences in the use of punctuation marks, morphosyntactic or syntactic differences, or differences in embedded elements such as index-markers. For a human being it takes only a short time to say that two sentences are similar.
Heyn: Translation Memories - Insights & Prospects
127
Spotting similarity using a computer, however, is another story. Cer tainly, classical computation based on binary oppositions cannot help us. Computation that does not rely on Boolean binary logic has tradition ally been called fuzzy. The problem with this term though, is that (in non-mathematical contexts at least) it is used very vaguely. Modem com puter science does, however, offer some workable solutions to similarity problems using fuzzy processing. These approaches include the use of neural networks and sparsely coded matrices. Whereas the first genera tion of the Trados translation memory system, for example, was based on a classical binary approach, and (linguistically motivated) substringoperations on classical database indices, the current generation employs sparsely coded matrices. The advantages are obvious: phenomena like misspellings and complicated syntactic deviations are now manageable and access time has been reduced significantly. Once a suitable tech nique for error-tolerant retrieval was introduced, additional functions like concordancing became possible. For the sake of simplicity, the term fuzzy-match is used in the CAT tools world to indicate the measure of similarity between two source trans lation units. It is important to understand that this is only a relative notion whereby a higher fuzzy-match value means more similarity. Interactive access
When users are accessing a translation memory in interactive mode, ac ceptable response times are in the range of up to one second. Response time depends, naturally enough, on the power of the computer, but also on the size of the translation memory, the type and number of processes running in addition to translation memory access, and the time spent in exchanging translation unit information between the translation memory system and the front-end (word processor). Translation memory size The size of translation memories is a real problem in systems with tradi tional data-access, i.e., those that do not use error-tolerant retrieval technology. The standard solution here is to extract from the master translation memory a smaller working translation memory which is made up of all stored translation units whose source segments are similar to the text to be translated. This is done in two stages. First the source text is segmented into translation units, according to the segmentation strate gies of the product. Then the translation units are retrieved from the translation memory and all matches above a certain threshold value are put into the working translation memory. The user then accesses this smaller translation memory interactively.
128
Unity in Diversity
Working translation memories, however, have a number of disadvan tages. For one, they necessitate an additional pre-processing phase. For another, the threshold similarity value is based on heuristics and users have no access to translation units below this value, even in cases where such units could be helpful. Also, if a translator changes the segmenta tion interactively, e.g. in case of segmentation errors, the system can no longer retrieve translation units. Interactivity is also compromised: the user is restricted to a buffer-translation memory and cannot work interac tively on the real translation memory. This means that changes to the working translation memory cannot be accessed by other users. Finally, the changes to the working translation memory must be resynchronized with the original master translation memory once the translation process is complete. This is again an additional processing phase and working translation memories only make sense if a versatile update mechanism is available to carry out resynchronization. In recent sparsely-coded-matrix based systems, real interactive work on ‘big’ master translation memories is possible. Big translation memo ries are typically in the order of 100,000 translation units, although memories in the range of 500,000 to 1,000,000 translation units are en visaged by the end of 1997. According to current research estimates, translation memories could be made up to 40% bigger without any in crease in constant access times. The ideal situation for all user groups is when a translation memory system is based on modem technology and at the same time allows for both the creation of a working translation memory and direct access to the master translation memory. Additional processes As already mentioned, response times can also degrade, sometimes sig nificantly, if additional processes, such as term recognition or the passing of translation units to an attached machine translation system, are run ning on the system alongside the main translation memory process. In order to avoid response time degradation, additional processes can be controlled by the translator or else performed in the background, so that the translation memory process always has the highest priority. Data exchange with the front end Normally, data exchange between the word processor and translation memory system is sufficiently fast. Batch P rocessin g
Batch processes are those that are carried out in non-interactive mode.
Heyn: Translation Memories - Insights & Prospects
129
The preparation of a working translation memory as discussed above is an example of such a process. Other batch processes are described below. Pretranslation Pretranslation can be carried out off-line. It involves the replacement of source text segments with all 100% matches and all fuzzy-matches up to a certain threshold, found in the translation memory. It may also involve the replacement of any source language terms detected by the terminol ogy recognition process. For ergonomic reasons, the system should highlight the results of the pretranslation process using a different colour. The advantages of pretranslation are that if a text contains many 100% matches, the translation can be performed much more quickly, since the translator can skip over the parts already translated, and that texts can be pre-processed and then be translated off-line by external users. As is the case with working translation memories, off-line translation requires so phisticated updating facilities in the translation memory system. Pretranslation is a process that is needed by nearly all user groups, in particular, translation agencies supplying work to freelancers. Repetition analysis Repetition analysis is a process that compares a document with a transla tion memory and computes statistics regarding how many 100% matches, fuzzy-matches and internal repetitions are encountered. In addition, word counts and translation unit counts as well as the overall statistical distri bution of items in a document can be output. Text segmentation must be sufficiently powerful to do proper word and translation unit segmenta tion and it must be able to cope with placeables correctly. Placeables are non-translatable items such as graphics and automatic field codes (auto matic numbering, dates, etc.) or tags, which are normally not translated, but simply p la c e d in the target translation unit by the translator. Repetition analysis is playing an increasing role in the negotiation of prices for translation projects. This means word, translation unit, and repetition counts must be correct. Where the repetitive character of docu ments is not easily measured (e.g. for users at the institutional level), delta computing, which allows the similarity of a set of documents to be gauged, offers a way of making estimates objective. Analysis of frequent occurrences of translation units The detection of all translation units occurring more than a certain num ber of times in a document can be very helpful. A list of these source translation units can be translated in isolation, thereby pre-filling the translation memory with the highly repetitive parts of the document. This
130
Unity in Diversity
is possible only if translation is feasible out of context, as is frequently the case for technical documentation. It may also be possible to export frequently occurring source translation units for which no matches above a certain threshold value can be found in the translation memory, to a machine translation system, in order to speed up the translation project. P ost-translation processes
Once translation has taken place, a final phase of revision and updating of the translation memory may be necessary. The translated text itself may also have to be cleaned up, especially in the case of source-preserving systems (see below). Updating and revisions The revision of translation memories is very important to all user groups, but especially when the reference and consistency factors play a major role. In interactive translation memory systems, updating is done auto matically by accepting a translation unit from the user. Simply re-opening a translation unit allows revisions to be done easily. This is the ideal situ ation since the updated translation unit is immediately visible to all users of the translation memory system. Users should also be able to update translation memories without using the front-end. This can be achieved by concordancing and editing the concordance results, that is by editing the translation memory directly. In non-interactive translation memory systems, and when using work ing translation memories, an explicit update has to take place in the form of a batch process when the translation project has been finished. Source-preserving systems A distinction should be made between systems that keep the source trans lation unit in the document in a hidden form and systems that do not. The first type of system, a source-preserving system, creates a bilingual docu ment in which the original source translation units are hidden. Such documents themselves contain a translation memory, and giving one away is like giving away a translation memory. Thus, once a translation project has been completed, all source translation units are usually deleted from the translated document. This is normally done by an update procedure. Although both source-preserving and non-source preserving systems allow working translation memories to be updated, only sourcepreserving systems are flexible enough for off-line revision. In off-line revision, revisers have access to source and target translation units in a document without using the CAT tool. This means that global replace operations and other facilities offered by word processors can be used
Heyn: Translation Memories - Insights & Prospects
131
completely independently of the translation memory system. In sourcepreserving systems, it is even possible to update a document from a translation memory. In cases where a translation memory is more up to date than a text, this can be very useful. Search-engine and data storage
There is often confusion about the distinction between the search engine and data storage in translation memory systems. The search engine is responsible for the retrieval of similar translation units and the data stor age is responsible for the physical storage of translation units. Physical storage can be done with any kind of database system. The architecture of the search engine is a more important factor in the performance of a translation memory system. As has already been pointed out, sparsely coded matrix approaches (a subtype of neural network) are currently state of the art, and significant improvements are not expected from tra ditional search engines, characterized in the case of translation memories by linguistically motivated string operations on data-storage indices. N etworking
The network capabilities of current translation memory systems also give rise to misunderstandings. Ideally, the following client-server scenario would prevail: if a large number of users were searching for different source translation units at the same time, a translation memory server would provide them with the required set of target translation units al most in real-time. So far there is no such system on the market. The only solution to this problem given a client-server architecture is to create a temporary working translation memory for each user, which is then cop ied to the user's workstation. File-sharing remains the only viable option for networked users. If a big group of users has to share a translation memory, a good choice is a system that fits major needs and allows development in the client-server direction. A dditional information stored in translation m em ory system s
To facilitate the interpretation of target translation units, they should be accompanied by additional formatting and administrative information. Administrative information must be user-definable: fixed field approaches are unacceptable. There must also be automatically maintained fields. Typically, these are fields such as creation date, creation user, change date, change user, used date, etc. Since these fields enlarge a translation memory, there should be a means of selecting or deselecting them. An automatically updated usage counter, which keeps track of the use of
132
Unity in Diversity
translation units allows for subsequent reduction of a translation memo ry to all the translation units that have been used at least once over a certain period. Users should also be able to select target translation units on the basis of a subject field code. Other possible fields include those used to install and enforce security mechanisms.
Front-end integration The term front-end refers to the application with which the translator controls the translation memory system. This is normally a standard word processor, but there are still some older systems on the market that come with their own editor. Such systems are, however, less than ideal, for a variety of reasons: firstly, the document has first to be converted into the internal editor format and later, after translation, it must be converted back into the original format. This can cause formatting errors and re quires additional work to be done. Secondly, idiosyncratic editors are not as user-friendly as standard word processors. They are character based and often lack automatic reformatting capabilities, multilevel undo/ redo operations, spelling checkers, thesauri, revision handling, auto-text entries, macro-languages and hyphenation facilities. Thirdly, translators are already familiar with their standard word processor, and it is a natural solution to integrate the translation memory into the environment transla tors are already familiar with. State-of-the-art translation memory systems are integrated into standard word processor systems like Microsoft Word for Windows or WordPerfect for Windows. Integration into an existing word processor can be done in several ways: the first distinction that has to be drawn concerns the dependence of the translation memory system on the word processor environment. Internal integration means that the translation memory system is com pletely integrated into the word processor using only the means for displaying and manipulation that the word processor offers. External integration means that the translation memory runs as an application independent of the word processor using its own windows to display re trieval results and its own menus for the manipulation of the translation memory system. Internal integration The advantage of internal integration is that the translation memory sys tem appears to end-users as a functional extension of the word processor. There are, however, several disadvantages associated with this approach: internal integration can use only the display facilities provided by the word processor. This means that there are clashes if a certain type of formatting is used within the document itself to mean one thing and by
Heyn: Translation Memories
-
Insights & Prospects
133
the translation memory system to mean another. For this reason, inter nally integrated solutions often avoid direct display, which means that users have to open and close windows in order to consult system infor mation. From an ergonomic point of view this is a major disadvantage. Internal integration also means a higher dependency on the word proces sor, which makes it more difficult - from a software engineering point of view - to integrate a translation memory system into new word proces sors re-using existing functionalities. Therefore, if internal integration is used, quick responses to new platforms or updates of word processors cannot be expected. External integration Externally integrated systems appear to users as applications in their own right. This has the disadvantage that users are forced to use a tool other than their tried-and-trusted word processor. On the other hand, transla tion memory systems integrate a number of functions, so that bundling all functions into one running application seems to be more natural than overburdening the already fully packed menu structures of modern word processors. External integration allows faster upgrades to new platforms, since only the part consisting of the communication with the word proc essor has to be re-implem ented. This is an im portant point to be considered when purchasing a system. All in all, and especially from an ergonomic point of view, external integration seems preferable. Indirect integration In some cases the front-end used for the creation of documents seems rather complicated to translators. This is the case especially with desktop publishing systems such as FrameMaker, PageMaker, Quark XPress or Interleaf. It thus often makes sense to convert from the desktop publish ing system to the standard word processor translators are familiar with. This type of integration is called indirect integration. Powerful conver sion tools have recently been developed, which smooth the complex format provided by desktop publishing systems into a format consumable by translators.1 The advantages of staying in the normal word processor environment outweigh the work involved in the two conversion steps. Depth of integration Systems differ with regard to the depth and sophistication of integration 1See especially the S-Tagger developed by itp-Ireland. This tool converts a Frame Maker file into a Microsoft Word treatable format, which translators can work on very easily.
134
Unity in Diversity
they allow. As already mentioned, a translation memory system identi fies a source translation unit and retrieves a similar or identical target translation unit. But what does the system identify as a source translation unit? A proper translation memory system should support all construc tions possible in the word-processing system, for example, tables, footnotes, endnotes, field-codes, frames, columns, embedded objects, pictures, indices, and revision codes. The segmentation capabilities of a translation memory system are also very important. Experience in computational linguistics has shown that segmentation is not at all a trivial task. There are ambiguities that cannot be resolved exactly (e.g. punctuation marks after numbers), lan guage-dependent phenomena (e.g. semicolon in Greek, abbreviations in Finnish, word boundaries in the languages of the Far East), and document-and user-type dependent phenomena (e.g. treatment of tabulars, semicolon, etc.), all of which cause segmentation problems. The only way to overcome these problems is to allow users to define their own segmentation rules, as well as lists of abbreviations, ordinal follow ers, etc. Some segmentation errors are, however, unavoidable. Therefore a system should allow the interactive shrinking and expanding of source translation units in order to specify exactly the size of a translation unit. If this is not possible, users will soon become dissatisfied. Automatic exchange of numbers and other invariable constructions is another useful function, especially in the area of banking, but users must be able to deactivate this function, if it does not apply to a certain document type. Segmentation must also foresee a means of exclusion. This means that parts of the document can be marked so that they can be excluded by the translation memory system. This must be possible at the paragraph level (e.g. to exclude foreign language citations or programming language code) and at the character level (e.g. to exclude invariable elements like proper names in biographical documents or tags in tagged file formats). Front-end independence
The storage of translation units in a translation memory should be inde pendent of the front-end. This means that, in an extreme case, a user can translate part of a document using, for example, Microsoft Word and the rest within WordPerfect, operating on the same translation memory. This is an important feature for the use of translation memories if different front-ends are used by different users, the user is planning a change of front-end, or translation memories are exchanged between different user groups. In principle this feature is valuable for all users. Front-end independence appears to be simple. Its technical realiza tion, however, is complex, because the formatting conventions of different
Heyn: Translation Memories - Insights & Prospects
135
front-ends have to be mapped into one single representation in a transla tion memory system. If a translation memory system offers front-end independence, this indicates highly sophisticated format management. A special front-end: concordance access
As has been mentioned before, concordancing allows access to transla tion memories. In this sense concordancing is a front-end of its own, enabling tasks like terminology searches and the maintenance of transla tion memories. We can expect to see many improvements in this area, for example, the use of filters for concordancing, the displaying of selective parts of a translation memory via concordance windows, and concord ance access to more than one translation memory. The next generation of the Translator’s Workbench, for example, will allow browser-like access to translation memories. Concordance access is important for all users, but especially for large institutions and terminologists. Only modem sys tems with sparsely-coded matrix technology allow for concordancing.
Emerging Issues A u thorin g
The use of translation memory technology can influence overall document production quite dramatically, often leading to a certain stream-lining in formatting. Unlike with machine translation, there is no need for a controlled language and the effect on overall text-flow organization is rather positive. The w ay translators see them selves
Mastering a technology can significantly improve the way translators see themselves. The use of modern computer technology influences nearly all professions. Being able to use new technologies and especially translation memory technology represents an additional professional skill for translators, one that is already highly appreciated on the market. P eep-h ole translations
The existence of translation memory technology may also influence the way translators formulate texts. For example, since retrieved translation units normally require fewer changes if they do not contain anaphoric and cataphoric references, translators are tending to avoid the use of such devices. The effect is a more technical style, and sometimes a less readable text. In the end it is up to the translator to decide whether text cohesion should be compromised in order to facilitate the translation memory.
136
Unity in Diversity
Translation rates
As has already been mentioned, the pricing of translations has been affected by translation memory systems. The localization industry already pays different rates for 100% matches, different types of fuzzymatches and no-matches, and other user groups are likely to follow suit in the near future. C opyright issues
As with term banks or dictionaries, translation memories can give rise to copyright problems. A translation memory can be extremely valuable as a source of terminology or a resource in retranslation. Ownership of a translation memory can guarantee an individual translator’s independ ence. Translation memories are thus valuable resources whose monetary value is very difficult to estimate. Copyright problems arise when it is unclear to whom a translation memory belongs, the supplier of a transla tion service or the client who commissions that service. In many cases this is subject to negotiation. Bilingual corpus collection
Translation memories are beginning to provide specialized multilin gual corpus material that is superior to automatically aligned corpora on two counts: its quality, because texts are translated manually and proof-read by specialists; and its domain and text-type specificity. Given fast (error-tolerant) concordancing, bilingual corpora can also compete with term banks.
The future Given that CAT tools providers are now seeking to support more lan guages, especially those of East Asia, we can expect to see developments involving UNICODE in the near future. Furthermore, although sim plistic approaches to the retrieval of subsegments of translation units using pattern recognition are showing astonishingly good results, such approaches can be applied only to certain types of texts. Future develop ments here will involve the integration of more linguistic knowledge and will therefore be restricted in the range of supported languages and the quality of retrieval for individual languages. As has been indicated above, work is also ongoing in the area of interactive client-server architectures for translation memories. One final area where things are changing rap idly is that of access to terminological information over the Internet. First applications in this area, such as MultiTerm for the WEB have already been released.
Consistency and Variation in Technical Translations A Study o f Translators’ Attitudes MAGNUS MERKEL NLPLAB, Department o f Computer and Information Science Linköping University, Sweden When large quantities o f technical texts are being translated manu ally , it is very difficult to produce consistent translations o f recurrent stretches o f text. One suggested remedy to this problem is to use tools based on translation memories. Successful use o f such tools presupposes that source repetitions can be transferred to the target text. This article reports on a study designed to validate this suppo sition. The study focuses on the distribution o f translations o f recurrent source sentences in two software manuals and uses a questionnaire to investigate how translators, project leaders in trans lation companies, and one customer evaluate variant translations in different contexts. The results show that translators mostly p re ferred consistent translations , but did not agree on which was the ‘b est’ translation. Project leaders and the customer also had con trasting attitudes to the questions o f consistency and variation.
Introduction When large quantities of technical texts are translated manually, it is very difficult to produce consistent translations of recurrent stretches of text, such as paragraphs, sentences and phrases. This can be due to many dif ferent reasons, for example, several translators work on different sections of the same document simultaneously, the source text is not final and may be changed at a later stage, and it may be too time-consuming or practically impossible to identify recurrent units in the source text manu ally. Individual translators making up a translation team will also have individual criteria for choosing a certain translation or even choosing from a set of possible translations. One suggested remedy to the problem of consistency in translation is to use tools based on translation memories. Successful use of such tools entails the supposition that source repetitions can be transferred to the target text. To validate this supposition, we first identified variant trans lations in two software manuals and categorized the variants. Then we designed a questionnaire based on the material and distributed this to the translators at the two translation companies who had translated the manu als. Apart from the translators, we also involved the project leaders at the
138
Unity in Diversity
translation companies and a representative from the customer (the soft ware company). The current study is a part of the project ‘Linguistic Engineering for Generation and Translation of Documentation’ at Linköping University (Ahrenberg & Merkel 1996).1 The study should be seen as complemen tary to assessments of practical usage of computer-assisted translation tools, e.g. Schaler (1994) and Vasconcellos (1994). To prepare the text material, we used our own tools for detecting recurrent sentences and phrases (Merkel et al 1994), and an alignment program based on a simplified version of Gale’s and Church’s algorithm (Gale and Church 1991). At the time of the study we did not have available a version of a verification tool which could pinpoint translation inconsistencies, so the variant translations used in the study were detected by manually scan ning sorted translation databases containing recurrent source sentences. However, since this study was completed we have constructed an auto matic tool that identifies such inconsistencies, cf. Merkel (1996). Although similar to TransCheck (Macklovitch 1994) in that it operates on bi-texts (i.e. representations of translations with explicit links between source and target sentences), our tool operates on consistent or inconsistent translations of sentences rather than lexical items.
Objectives and methodology This study focuses on the distribution of translations of recurrent source sentences and how translators familiar with the text type evaluate the existing translations of the recurrent units in different contexts. The re sults of the questionnaire provide important insights into the usability of translation memory tools by showing, for example, what types of recur rent segments can be translated uniformly, and what kind of variant translation is preferred. The questionnaire was divided into two parts. The first part involved nine questions regarding translators’ attitudes towards the text type, their potential uses of various translation tools or techniques, and their atti tudes towards the benefit of using such tools. The second part consisted of 50 examples from two computer manuals from the same company, where each example described an identical source segment (paragraph, sentence, heading, etc.) occurring in two different contexts. These source segments were shown with the corresponding target segment, which had 1This project is financed by the Language Technology Program jointly managed by the Swedish Council for Research in the Humanities and Social Sciences (HSFR) and The Swedish National Board for Industrial and Technological Development (NUTEK).
Merkel: Consistency and Variation in Technical Translations
139
been translated in two different ways. The task for respondents was to decide whether they would prefer consistent translations in the two con texts or the two variant translations. If consistent translations were preferred, the respondent was asked to motivate the choice and also rank the different alternatives. If variant translations were preferred, the re spondent was asked to justify this choice. The questions accompanying each example were identical. To avoid any influence caused by the way the examples were ordered, we randomized the order of the 50 examples for each respondent. The questionnaire was sent out to both in-house and freelance transla tors working for two Swedish translation companies. One requirement was that translators should be familiar with the text type and the require ments of the translations (i.e. that they should have translated texts from the particular software company before). We also asked a project man ager at each of the translation companies to complete the questionnaire as well as the person responsible for translation quality at the client soft ware company. In total we received 13 completed questionnaires from translators (8 from freelance translators and 5 from in-house translators). We also received feedback from the two project managers and the client software company. To distinguish between different categories of translation variants, the examples used in part two of the questionnaire were divided equally (five examples from each category) over the 50 examples, as shown in table 1. We distinguish between three major types of translation varia tion: synonym variants (where the variants have the same underlying logical form), partial synonym variants (where the variants differ in de gree of specification) and non-synonym variants (where the variants do not have the same underlying logical form). We also include a fourth category where longer stretches of text were repeated (category D be low). The numbers in brackets indicate the example numbers used in tables 2 and 5.
Results and implications Nine questions were included in part one of the questionnaire. The ques tions involved the respondents’ general attitude towards consistency and variation, translation tools, tools to identify and translate recurrent para graphs, sentences and recurrent fuzzy patterns (such as From the X menu , choose Y.) We asked if they used the built-in search-and-replace func tion of word processors while translating. We also asked them for their opinions regarding translation memory tools and a hypothetical transla tion verification tool that highlighted variant translation. Finally, we enquired if they had any other comments on translation tools.
140
Unity in Diversity
All translators agreed that terminology was important. Twelve out of thirteen also stated that sentences and phrases should be translated con sistently and that the source text often shows too much variation. Six translators found that it was difficult to know what was actually recurrent in the source text, especially across chapter boundaries. Two translators wanted variation of recurring sentences in running (descriptive) texts, but not in instructions. Still, the majority of the translators did not see variation as a goal in itself. One translator expressed this as “Variation is only confusing. The text is probably not read from cover to cover, instead the reader looks up different things.” All translators use term lists in electronic format and all of them use the printed guidelines issued by the software company. Twelve out of thirteen translators use term/word lists in printed form as well, but only two use look-up facilities directly from the word processors. None of the translators used translation memory tools for these particular texts, but two translators used such tools for other customers’ translations. Two translators claim that they use tools to identify and translate re current paragraphs. Five translators use the search-and-replace function in the word processor to do this manually. The search-and-replace func tion is used when the translator notices that certain segments are repeated during translation. Six translators do not use tools for this purpose. No translators have access to tools which automatically identify close matches (fuzzy matches). Two translators state that they use search-andreplace with wildcards occasionally. Eleven translators use the built-in search-and-replace function of their word processor regularly; however, many of them added that they do this “with caution” or that they “always check replacement case by case”. Two translators state that they do not use the search-and-replace function at all. Nine translators have a positive attitude towards the use of translation memories; four show some degree of hesitation. The reservations made are that translation memories “must be easy to use”, that they should be used “only as a reference”, and that translators “have to be careful be cause terminology may have changed”. One translator expresses the fear that it could be “tempting to work too quickly which will lead to an in creased number of mistakes”. No translator was totally against the use of translation memories. Nine translators are positive towards a verification tool, three have some doubts and one translator sees no use for this kind of tool. One translator noted that such a tool would be especially useful for checking terminology and expressions. Typical comments from the translators dealt with concerns about the quality of the source text. Some also expressed a fear that translation work will become more tedious and boring, and that some of the creative
Merkel: Consistency and Variation in Technical Translations
141
aspects of the job will disappear with the increasing use of translation memory tools. Some translators expressed concern over the changing role of the translator, from a linguistic innovator to a linguistic operator, or as one of the respondents phrased it: “The translator is reduced to some body who presses the OK button.” Table 1. Variant translation categories M ajor type of variation
Variation categories
Examples
la. Syntactic variants same context types (No. 1-5)
Running text-running text, heading-heading, etc.
lb. Syntactic variants different context types (610)
Running text-heading, table cell-heading, etc.
2a. M orphem atic variants - same context types (1115)
Running text-running text, heading-heading, etc.
2b. M orphem atic variants - different context types (16- 20)
Running text-heading, etc.
3. L exical variants (2125)
N on-term inological syno nyms
4. Coherence variants (26- 30)
Pronouns, adverbs, etc.
B. Partial synonym variants
5. Specification variants (31- 35)
Less or more specific content
C. Non-synonym variants
6. Term inological variants (36-40)
Varying term inology
7. Content variants (4145)
Erroneous content translations
8. More than one recurring subsequent sentence (46-50)
Repeated paragraphs or whole sections
A. Synonym variants
D. Recurrent multisentential segments
Translation companies9 attitudes The questionnaire given to project leaders at the two translation compa nies was similar to part one of the translators’ questionnaire, but it was geared towards what the project leader wanted the translators involved in the project to do. In translation company A, the project leader is positive towards trans lation tools. At translation company A, they do not use tools for handling
142
Unity in Diversity
recurrent paragraphs or sentences in general, but the use of search-andreplace functions is encouraged, if it is done using good judgment. The project leader thinks that the use of translation memories would benefit the company, and the same applies to the use of a verification tool. She states that tools would make translations of references to chapter and sec tion headings, for example, easier in printed documentation and in on-line help texts. However, she concludes, when running text is involved, there is a risk that the creativity and ability to localize text can be inhibited. In translation company B, the project leader is, in general, very posi tive towards translation tools. Specifically, he would like the translators to use search-and-replace functions. At translation company B, they have created their own tools for handling recurrent sentences, as well as their own terminology tools. He also thinks that the translation memories and verification tools would be helpful to translators and editors. However, he emphasizes that it is important that translation tools be very flexible and easy to use. Furthermore they must include functions for handling up-dates of the source text and provide room for a review and verifica tion phase.
The customer’s attitudes In an interview with the person responsible for the Swedish translations (terminology and quality assurance) at the customer company, the fol lowing views were expressed. Translation memory tools are an excellent help for repetitive texts. The prerequisite is that the previous translation memories are correct, which is not always the case. Terms may have been changed and there may be mistakes in the old translations. In general it is not advisable to reuse an old translation without verifying the accuracy of the transla tion. This verification must be done before a translation memory can be used in a new project. It is important that translators be able to view the context in which a certain sentence is to be translated. A pure sentenceby-sentence translation is not advisable at all. Automatic translation without the confirmation of the translator is not acceptable. From the questionnaire, it was clear that the representative for the customer was less inclined to prefer consistent translations than the trans lators. In the interview, this opinion was explained by the fact that the objectives may be different for a customer and a translator. The custom er demands a high quality translation and is not really interested in how this is achieved, whereas the translator is by definition involved in the actual process of translation and strives to minimize the effort of reach ing high quality translation. This may be the cause of the discrepancy shown in the questionnaire, i.e. that the customer representative had a much higher proportion of “doesn’t matter” replies than the translators in
Merkel: Consistency and Variation in Technical Translations
143
general. For example, for running text the customer representative stated that different translations were possible, but not necessary, whereas the translators preferred consistent translations to a higher extent.
Consistency vs. variation As described earlier, the respondents were asked to state whether they preferred consistent translations of a given source segment in two differ ent contexts. The options given were either yes or no , with a space for the respondent’s own motivations for his/her choice. When I examined the questionnaires, it became apparent that there was a need for a third re sponse, in between yes and no, namely a response which we can call “doesn’t matter”. This applies when the translator in the justification for the choice has indicated that the translation could be consistent, but that it would not matter whether the source segment was also translated differ ently. This introduction of a third response category may seem to diminish the possibilities for interpreting the results, but if we still regard the re sponses as binary (yes or no), with the judgement of consistent translations as the primary problem, the responses “yes” and “doesn’t matter” both indicate the possibility of consistent translations, whereas a “no” response rules out consistency. Table 2 summarizes the results from the question naire regarding the respondents’ choice of whether a source sentence should be translated consistently or not. The “YES” option shows how many respondents judge that the source sentence should be translated consistently. The “NO” option shows how many respondents want dif ferent translations in the two contexts and the “DM” (Doesn’t Matter) option states how many respondents regard it possible, but not necessary, to be consistent. A first glance at the answers indicates that there is a clear preference for consistent translations; for 48 examples a majority of the respondents prefer consistent translations. Furthermore, in 38 of the examples all re spondents have ruled out variation (zero value for NO). In two of the examples, a majority of the respondents are in favour of variant translations (examples 7 and 29). The first is an example where Calculating with precision as displayed: occurs as both a heading and as cell item; that is, in different contexts. The second (example 29) depicts the source sentence This value is derived using the form ula:, where the first translation is spelled out with a definite description and the second has kept the coherence marker (th is valu e) of the source sentence (“Pearson-korrelation beraknas m ed foljan de f o r m e r vs. “D etta varde beraknas m ed foljan de fo r m e r ).
To compile the preferences related to categories, we have added the number of responses in each category and distinguished two major classes:
2
3
4
5
6
7
8
9 70 77 12 13 14 75 16 77 18 79 20 27 22 23 24 25
1
0
6
8
0
0
NO
0
4
0
2
DM
4 5
8
0
1 0
1
2
2
12
3 0
0
4 0
2 0
2 0
5
6
0
6
0
1 1 0
3 2
3
2
1
4
4 5 0
0
0
0
0
0
2 0
0 0
0 0
0 0
1 0
0 0
0 1
3
0
1
0
0
0
1
0
0
0
1
0
3
0
3
0
1
14 16 16 16 15 15 16 12 15 16 15 16 15 13 13 13
0
2
Table 2. Responses to questions on consistency
0
1
15 15 15 15
14 16 12 2
YES
0
1
26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
0
0
Be No
0
1
2
1
0
0
4
NO
1
2
2
2
14 13 14 15 12 16 6 10 12 25 14 12 14 14 11 10 10 11 11 11 13 14 13 14 15
7
DM
YES
No
Ex
144 Unity in Diversity
145
Merkel: Consistency and Variation in Technical Translations
categories with a clear preference for consistency and categories where the respondents have shown a marked degree of hesitation towards con sistency. Tables 3 and 4 show the number of responses and the number of examples over which the response is distributed. Table 3. Eight categories with a preference fo r consistency C ategory
YES
D o e sn ’t M atter
NO
la . Syntactic variants - sam e contexts
68/5
11/5
1/1
2a. M orphem atic variants - sam e contexts
65/5
15/5
0/0
3. L exical variants
69/5
11/5
0/0
4. C oherence m arkers
59/5
13/4
8/1
5. D egree of specification
71/5
7/5
2/1
6. T erm inological variants
79/5
1/1
0/0
7. C ontent variants
74/5
5/3
1/1
8. M ulti-sentential variants
70/5
10/4
0/0
In category 4, there is a single example (number 29) where the re spondents actually preferred variant translations, but this seems to be an exception as there was a clear preference for consistency in the other four examples. (Example 29 is discussed above.) Table 4 . Two categories with a marked degree o f hesitation fo r consistency Category
YES
D oesn’t M atter
NO
lb. Syntactic variants - different contexts
60/5
4/3
16/3
2b. M orphem atic variants - different contexts
58/5
2/2
20/4
In table 4, the figures for the two categories where the functional con texts differed are shown. This contextual parameter seems to be the deciding parameter. There is still a majority of YES answers, but the number of NO answers is considerably higher for these categories. The trend for the categories with different functional contexts is definitely more towards the ‘hesitation’ side than in other categories. The responses here can only be interpreted in one way, namely, that consistency is something that technical translators aim for in general. The exception to this rule regards context. When two source sentences (or segments) occur in different structural contexts, such as headings and table cells, translators should be more cautious in applying consistent
146
Unity in Diversity
translations. These recurring source sentences may often require differ ent target sentences. In some translation software this is indeed handled by recognizing that they have different formatting tags (styles or other mark-up properties) which means that no perfect match will occur, only a fuzzy match which leaves the translator to make the choice of whether or not to be consistent.
Uniformity when deciding on a preferred translation alternative In the previous sections, we have seen that there is an overwhelming general tendency towards consistency. One side-effect of the question naire was that we received a great deal of data on how uniform the choice of alternative was between different translators. As described earlier, the translators were asked not only to state whether they preferred consist ency or variation in the translation pair given, but also to state which of the alternatives they preferred. They were also asked to say whether the alternatives were equally appropriate or whether they preferred another translation. Given each example, the translator had four choices: (a) P re f e r A , (b) P refer B , (c) Prefer A o r B (equally appropriate), or (d) Prefer other translation.
As the respondents were presented with only a limited context (around two paragraphs), we expected a certain degree of disagreement in the respondents’ answers. However, the differences were far greater than anticipated. Table 5 summarizes the choices from the questionnaires. A indicates the number of respondents who choose alternative A for each example, B alternative B, ‘Other’ indicates ‘other translation’ and EQ suggests that both A and B are equally appropriate translations. For each category, the alternative with the highest number is indicated in bold face. Note that the A and B alternatives do not signal any ‘degree of appropri ateness’ on our part, they are chosen completely randomly. In only one example (32), one of given the translation options is ruled out entirely, i.e. no translator has chosen alternative A in example 32. Even though some examples contain zero values for A or B, these zero values are offset by the non-zero value for EQ (in examples 3, 14, 17, 21, 36, 38, 40 and 48). We asked the respondents to state motivations for their choices and when we studied the responses we found that most of the motivations given were vague and subjective. There are numerous ‘feelings’, and value statements such as ‘better’, ‘clumsy’, ‘correct’ and so on. Some prefer translations closer to the source, some want them to be more ‘Swedish’. Some prefer more specific and some prefer more general translations. In several cases, we found totally contradictory motivations for the same example. Many respondents do not give motivations at all, and the ones
11
2
1
2
A
B
Other
EQ
3
6
0
0
1
0
3
5
2
12 7
1
6
0
6
4
5
3
3 1
2 4
0 3
3 4
5
0
1
2
3
2
1
2
9
4
0 3
1
5
1
0
3
1
8
10 3
8
0
4
5
7
0
8
0
12 6
4
Table 5 . Preferred translation alternatives
0
1
5
1
1 15 11 7
2
2
5
0
4
0
3
9
1
0
8
0
4
2
6
0
5
1
4
5
5
2
1
4
2
1
4
0
11 3
1
3
1
3
2
3
1
2
5
1
5
3
0
8
1
0
3
6
1. 2
1 10 7
14 2
4
0
5
2
2
2
7
5
8
2
5
1
5
7
0
4
3 3
9
9
1
6
2
1
4
6
1
7 2
6
1
1
1
3
8
6
26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 4S 49 59
4
1
2
2
2
0
4
Ex
4
1
6
4
2
5
4
8
6
1
4
4
3
EQ
1
3
8
2
10 11 12 13 14 15 16 17 18 19 20 21 22 2? 24 25
2
4
3
1
9
Other
11 10 5
8
7
1
6
6
7
6
5
B
1
4
1 10 0
3
A
2
1
Ex No.
Merkel: Consistency and Variation in Technical Translations 147
148
Unity in Diversity
who do are very brief. Perhaps the motivations would have been more elaborate if this aspect of the questionnaire had been emphasized, but then some translators may not have filled in the questionnaire as it would have required considerably more time to complete. The lack of consensus when picking the ‘best’ translation gives rise to certain questions regard ing the use of translation support software. If these translators had worked using a translation memory-based tool, and if one of the alternatives (A or B) had been presented as the suggestion to use, what would their reaction have been? Previously we have concluded that the general aim is to strive for consistency in these kinds of translations, given certain constraints on context. Would they actually use the suggested alternative or would they prefer something else? Something that is ‘better’ or ‘more Swedish’, for instance?
Summary Translators of software manuals do strive for consistency in general. On the whole, they have a positive attitude towards translation tools, such as translation memories, but show some hesitation regarding the change of the translator’s role when working with such tools. The only explicit reason for not choosing consistency is when a repeated source segment occurs in different functional contexts, for example as a heading and as a cell in a table. The choice of ‘best translation option’ varies considerably among translators, which indicates that it may be difficult to encourage transla tors to accept suggested translations from translation memory-based programs. The trend seems to be that the customer is not as negative towards variation in running text as the translators and project leaders. One rea son for this may be found in the different perspectives that the three types of participants have in the translation project. However, to make this a substantial claim, a larger study involving several translation companies and customers would be necessary.
References Ahrenberg, L. and M. Merkel (1996) ‘On Translation Corpora and Translation Support Tools: A Project Report’, in K. Aijmer, B. Altenberg and M. Johans son (eds) Languages in Contrast: Papers from a Symposium on Text-based Cross-linguistic Studies , Lund: Lund University Press, 185-200. Gale, W. and K. W. Church (1991) ‘A Program for Aligning Sentences in Bilingual Corpora’, in Proceedings o f the 29th Annual Meeting o f the Association fo r Computational Linguistics , Berkeley, CA, 177-184.
Merkel: Consistency and Variation in Technical Translations
149
Macklovitch, E. (1994) ‘Using Bi-textual Alignment for Translation Valida tion: the TransCheck System’, in Proceedings o f the First Conference o f the Association fo r Machine Translation in the Americas (AMTA-94),
Columbia, MD, 157-168. Merkel, M. (1996) ‘Checking Translations for Inconsistency - A Tool for the Editor’, in Proceedings o f the Second Conference o f the Association fo r Machine Translation in the Americas (AMTA-96), Montreal, 157-167. Merkel, M., B. Nilsson and L. Ahrenberg (1994) ‘A Phrase-Retrieval System Based on Recurrence’, in Proceedings o f the Second Annual Workshop on Very Large Corpora , Kyoto, 99-108. Schäler, R. (1994) ‘A Practical Evaluation of an Integrated Translation Tool during a Large Scale Localisation Project’, in Proceedings o f the Fourth Conference on Applied Natural Language Processing (ANLP-94), Stutt gart, 192-193. Vasconcellos, M. (1994) ‘The Current State of MT Usage Or: How Do I Use Thee? Let Me Count the Ways’, The LISA Forum Newsletter 111:21-29.
Page Intentionally Left Blank
The Problem with Machine Translation REINHARD SCHALER Localisation Resources Centre University College Dublin, Ireland Recent studies and reports have shown that Machine Translation (MT) is about to by-pass the translation profession. In this article we will argue that this development is due to the translation profes sion's inability to resolve the tension between its traditional professional values and reference system, and the new objectives o f translation technology. We hold that a change in the translator's professional mind-set has become necessary if translators do not want to exclude themselves from some o f the most interesting and lucrative areas o f translation activity, and suggest that this can be achieved through: the integration o f translation technology at all levels o f translation studies courses; the establishment o f Transla tion Technology Centres; and the provision o f better financial and political support fo r join t industrial and academic research projects at national and European level.
As with any other software product, there are many problems associated with Machine Translation (MT) systems. Few software products, how ever, have been so harshly judged by their potential users as MT systems have been. Translators widely agree that “Machine Translation does not work”. They maintain that translations produced by currently available commercial MT systems are generally of unacceptable quality. They con sider that the time and the costs involved to improve this quality to a level where publication could even be considered make it redundant and are convinced that they, the human translators, produce better quality trans lation at the same speed as MT systems and at a lower overall cost.
Translation as an essentially human activity Over the last few decades translation theory has been introduced at poly technics and universities around Europe. This discipline deals with the body of knowledge we have on translating and sets out to provide trans lators with a frame of reference for translation and translation criticism. One of its aims is to identify and define a translation problem, to indicate all the factors to be taken into account when searching for a solution, to list possible translation procedures, and, finally, to recommend the most suitable one together with the appropriate translation. The theory claims to be based firmly on problems arising in translation practice, mainly the problems of naturalness (both grammatical and lexical) and the
152
Unity in Diversity
relationship between language and reality. One of the most widely recommended textbooks in the Englishspeaking world on translation theory is Peter Newmark’s A Textbook on Translation. Newmark’ s book appears on the reading list of many courses for professional translators and contains a very well structured, compre hensive and widely accepted coverage of the most important issues in translation. In his summary, Newmark (1988:224-225) gives six reasons why trans lating can be so enjoyable and satisfying: 1. Translators are explaining something, pursuing the subtleties of ideas. 2. Translation is a continual chase after words and facts with success depending entirely on the translator (though the element of luck is important) and rewarded by the joy of finally finding a word in a book after hours of searching on the shelves and in one’s mind. 3. It is never-ending, because a translation can always be improved, be cause it gives a tactile feeling and relishing for words as well as the rhythms of sentences read aloud to oneself. 4. The challenge, the wager, the isolation - often translators write on behalf of an author they do not know to readers they never meet. 5. The joy of the find, the happy concise stretch of language, when trans lators feel they have written just what the author wanted to but did not. 6. Because of the sense, when translators are translating some novel or biography, that they are identifying not only with the author but with the main character, and incidentally with someone dear to them who appears to embody them. Newmark undoubtedly has captured the essence of what translation means to many translators who see their work as a creative activity pro viding great intellectual satisfaction. For many professional translators, translation is just another stage of writing, taking a step beyond, adding a vital new dimension to the original. This approach is echoed in conference papers and discussions among translators, who regularly debate with great passion issues like whether translation is a process that should transform the alien into the familiar and mould the original to fit a comfortable cultural framework or whether it should do the opposite: acknowledge the foreignness or alien quality of the original in the translation. One of the best known Irish translators, Gabriel Rosenstock, who has translated the Irish poet Seamus Heaney into Irish, compares the art of translation to a blood transfusion between friends. It is obvious that translators who have been educated and trained fol lowing a syllabus such as that proposed by Newmark and who have consequently acquired an enormous sense of responsibility to care for
Schäler: The Problem with Machine Translation
153
their language and a pride in their work cannot and should not be down graded to MT operators. Translators do not like subservience to a machine, they do not want to revise the poor quality output of MT sys tems (Hutchins 1995). However, even translators have to earn a living and will eventually have to adjust to the fact that MT has become a reality in many commer cial environments. SAP, Europe’s biggest software developer, for example, has a staff of 8 full-time employees in its MT Service Group with a turn over of 500,000 words per month. They report high-quality, high-speed translations with an excellent profitability rate (Grasmick 1995). Other companies, especially those involved in the localization of software like Lotus Development, Oracle and Corel, have only recently become MT users or are actively considering the introduction of MT systems. Given its traditional values and reference system, the translation pro fession is currently not in a position to cope with the requirements and new objectives of translation technology.
Translation without translators Recent studies and reports have shown that - because of the introduction of MT - translation is no longer the exclusive domain of translators and non-translators are the biggest users of MT systems. The European Commission is one of the biggest MT users in Europe. Their MT system, Systran, is available via an internal electronic mail network to all Commission officials. A recent in-depth study of the Commision’s use of MT revealed some remarkable figures (Senez 1995): • • •
Between 1988 and 1994 there was a 35-fold increase in the use of Systran (from 4,000 pages to 140,000 pages) MT is being used by about 2,500 staff (30% of those are ‘regular’ users, i.e. they use the system at least 5 times a month) 20% of the users are in the Translation Service, 80% in other Com mission departments, i.e. the majority of users are non-linguist staff who avail of machine translation when the need arises.
Another indication of the appeal of MT to non-translators is Compu Serve’s use of the Intergraph MT system in its World Community Forum, which attracted some 15,000 subscribers in less than three months of op eration (Brace et al 1995). Furthermore ‘toy’ MT systems, i.e. systems not intended for pro fessional translators, are phenomenally successful while systems designed for professional translation struggle for survival. The latter are mainly installed in translation companies staffed by professional translators who - because of their educational background and professional experience -
154
Unity in Diversity
are generally reluctant to work with MT and to take on the new role of MT coder, pre-editor or post-editor. On the other hand, MT systems like the PC-based Power Translator and Translation Assistant series, consid ered for a long time to be for the home user only and not suitable for business applications, have been released as new Windows product lines with all the characteristics of the ‘mature’ systems (Brace et al 1995). By the end of 1994 they had sold more than 400,000 units - a figure 1,000 times higher than that predicted by an OVUM report in 1991. The exact number of ‘high-end’ installations like Systran, Metal and Logos is not known, but it is currently estimated to be about 4,000 times lower than the number of ‘toy’ units quoted above and such high-end systems have a very low number of active and regular users. Few high-end MT users use systems to their full capacity in an attempt to improve the qual ity of MT output. This is especially true when companies have to deal with short turn-around cycles and frequent updates of technical texts, and when accuracy and consistency of the translation (features associated with MT output) take precedence over style, readability and natural ness, all associated with the traditional values and reference system of human translators (Tools Group Papers forthcoming). Long-standing industrial users of MT say they will not hire translators for their MT departments and advise MT novices to follow suit. One of the most experienced users of MT, the National Air Intelligence Center (NAIC) in the United States of America, who recently celebrated over 30 years of operational use of MT, does not hire translators but computer scientists, post-editors and language specialists (Bostad, personal com munication 1995). According to the Center most translators actually spend more time trying to prove that the MT system ‘does not work’ than trying to ‘co-operate’ with the system in order to produce acceptable out put. The NAIC thus prefers to employ linguistically less qualified staff as pre- and post-editors, e.g. subject specialists with a knowledge of the source language, in order to ensure the smooth and cost-effective opera tion of their MT installation. No reliable figures could be obtained to size the world market for translation, but it is estimated that around 250 million pages of techni cal and commercial text alone are being translated every year. Although MT is currently only used to translate a fraction of this, around 1.2 mil lion pages per year (Vasconcellos 1995), the potential for MT is obvious. Many Irish based companies involved in software localization, for exam ple, are currently installing MT systems and it is expected that this trend will continue.
Building bridges: Machine Translation with translators The problem with Machine Translation - from the translator’s point of
Schäler: The Problem with Machine Translation
155
view - is not that ‘MT does not work’, i.e. that an enormous amount of work still has to be done to improve commercially available MT systems. This is the problem of MT developers. Nor is it simply a problem of translators having the wrong attitude towards MT, as some ‘experts’ sug gest. We also oppose the view that translators are basically a hindrance to the successful running of an MT operation. However, it is evident that the translation profession has to resolve the tension between its traditional professional value system and the new technologies if it wants to participate in some of the most interesting and lucrative areas of translation activity. This requires radical changes in the translator’s professional mind-set, a process which has to be supported by clear and decisive measures. The teaching of translation technology has to be integrated into all levels of translation studies courses. At the end of their professional edu cation, translators must be aware of the wide variety of translation tools available, including MT, and have had some exposure to a representative selection of these tools. They should be able to specify requirements for MT systems, and to draw-up criteria for the evaluation of these systems. In addition, they should have learned about the financial and operational implications of the introduction and use of MT and other translation tools in a traditional translation environment. Translation Technology Centres should be established, in order to make state-of-the-art translation tools more accessible. Initiatives like the establishment of the Localisation Tools Library at the recently launched Localisation Resources Centre in Dublin are a step in the right direction; the high level of support they have received from indus try and the translation profession is an indication of how much such initiatives are needed. Better financial and political support at national and international level has to be provided to encourage more joint industrial and academic research projects in very practical areas like the evaluation of MT sys tems. Such evaluations should be based on properly researched user requirements rather than on academically inspired theoretical foun dations. Projects like the Evaluator’s Workbench, proposed to the European Commission under the 4th Framework Programme, are a good example of pan-European co-operation between industry and research ers in this area.
Conclusion Translation and machines are not mutually exclusive. But more than one bridge remains to be built to span the gulf between language engineers and translators. We must ensure that all aspects of translation activity remain under the control of the translation professional and are not taken
156
Unity in Diversity
over by highly qualified engineers with no feel for and no knowledge of the human element in natural language. Translation, even Machine Trans lation, is not possible without translators. But it is incumbent upon translators to broaden their concept of translation and to ensure that their skills are recognized as fundamental to all types of translation activity.
References Brace, C., M. Vasconcellos and C. Miller (1995) ‘MT Users and Usage: Europe and the Americas’, in M T Summit V, Proceedings, European Commission and International Association for Machine Translation. Grasmick, D. (1995) ‘Two MT Systems and still Hungry’, in MT Summit V, Proceedings, European Commission and International Association for Machine Translation. Hutchins, J. (1995) ‘Reflections on the History and Present State of Machine Translation’, in M T Summit V, Proceedings, European Commission and International Association for Machine Translation. Newmark, P. (1988) A*Textbook o f Translation, New York & London: Prentice Hall. OVUM (1991) Natural Language Markets: Commercial Strategies , Report compiled for Ovum Ltd. by B. Engelien and R. McBride, London. Senez, D. (1995) ‘The Use of MT in the Commission’, in M T Summit V, Proceedings, European Commission and International Association for Machine Translation. Tools Group (forthcoming) Report o f the Tools Group o f the Software Lo calisation Interest Group SLIG , Dublin: Localisation Resources Centre.
Machine Translation as a Model of Human Translation PAUL BENNETT Centre fo r Computational Linguistics UMIST, Manchester, UK Machine translation (MT) has generally been seen as a purely en gineering enterprise, with virtually no attention paid to whether it provides any kind o f psycholinguistic model o f human translation (HT). This article investigates whether M T systems offer any kind o f parallel to HT, by distinguishing three ways in which MT may simulate HT, namely in terms o f input-output relations, knowl edge, and processing. It is suggested that M T does not, in general, produce human-like output. And while the knowledge embodied in an M T system is broadly comparable to that o f an expert human translator, the form er is more compartmentalized, and specifically bilingual knowledge is sparser. Processing is hard to discuss without knowing more about how people translate. However, MT can play a useful role in prompting hypotheses about HT such as whether there is a human analogue o f M V s complex transfer.
Introduction Machine translation (MT) has generally been seen as a purely engineer ing enterprise, designed to produce translations of acceptable quality, rather than as an attempt to build a psychologically accurate model of the way in which human translators work. In this article, however, I wish to raise the issue of the psychological status of MT, to ask whether, and in which ways, current MT systems simulate human translation, and whether this is a useful design criterion. I am interested, therefore, in the extent to which MT research can be seen as part of Computational Psycholinguistics in addition to Applied Computational Linguistics (see Thompson 1983). I shall begin by setting out three different respects in which MT might be said to simulate human translation, and then discuss each in turn. The tentative and preliminary nature of this enquiry should go without saying.
Kinds of modelling I distinguish three ways in which a computational system might be de scribed as a model of some aspect of human behaviour: ( 1) a. b. c.
input-output relations knowledge processing (use of knowledge).
158
Unity in Diversity
In the case of (la), a system behaves in the same way as a human as far as its inputs and outputs are concerned, i.e. it relates a source and target text. An MT system which achieved this would produce humanstandard translations, or at least would produce translations. This kind of mimicking of translators’ behaviour may seem to be a trivial kind of modelling, though I shall suggest that it is not in fact trivial. (lb) requires a system to represent, in some form, the knowledge of a human practitioner. For an MT system, this would mean at the very least that it embodied the kind of monolingual and bilingual knowledge, both grammatical and lexical, possessed by a translator. One might require further that this knowledge be structured in a way that paralleled that of the human, though clearly even this leaves a great deal of leeway. The final kind of modelling, (lc), implies furthermore that the knowl edge be used or exploited in a comparable way to that of a human. In our case, the process of translation would follow the same course in the per son who translated and in the computational model of translation. I take it that a system that was a full-scale model of the human translator would produce human-quality translations, but I regard it for now as a moot point whether aiming at (lc) would be the best way of pursuing better translation quality in MT. Having established in broad terms the kind of modelling I am con cerned with, I shall now ask just how an MT system could be said to be a model in any of these three senses. Before that, however, I want to men tion a reason why questions of psychological reality might well be seen as irrelevant to MT. Translation differs fundamentally from parsing in being a conscious, learned activity, not an automatic, unconscious reflex (as pointed out by Johnson 1983). Translation is not an input system in the terms of Fodor (1983), and is thus quite unlike monolingual language processing and vision, two principal areas of computational modelling of human behaviour, and there would be little reason to expect it to be ame nable to comparable study. The processes of pre- and post-editing, which play an important role in most MT applications, also have little place in ‘ordinary’ language processing. There are other reasons, too, why one might balk at such compari sons. For instance, Sager (1994:258) argues that a machine-translated text is simply not comparable to one produced by a human. In my view, however, such comparison is permissible provided that one bears the limitations of the procedure in mind and does not use it for the purposes of contrastive evaluation.
Modelling input-output relations With regard to (la), I claimed in Bennett (1993a) that there is no point in building an MT system which makes the same mistakes, or the same kinds
Bennett: M T as a Model o f Human Translation
159
of mistakes, as human translators. If the aim is the best quality translation achievable, it makes no sense to build in to the system deliberate mistranslations. I would now like to qualify this view, however, by not ing that it would be interesting if MT systems made human-like mistakes without this having been specifically intended. In a comparable way, the parser for English written by Marcus (1980) stumbles over garden path sentences such as The horse raced p a s t the barn fe ll, just as people processing such examples do. This leads us on to the more general issue of the extent to which the output of MT systems resembles expert human translations. Given the number of MT systems and human translators around, one might be re luctant to speak of ‘typical’ examples of either. However, I think it would be generally accepted that even good-quality MT systems do not on the whole produce human-like output, tending to produce literal or structurebound translations (see Hutchins & Somers 1992:138; Bennett 1994). Systems specialized for particular domains and text-types perform bet ter, of course. Where a translation produced by a machine is inadequate, it is often ill-formed in terms of the target language grammar, which hu man translations generally are not, so there is little case for psychological modelling here. To illustrate this last point, consider two examples of Russian-English translation by SYSTRAN. These are taken from Knowles (1979); in each case, I give the system’s output followed by a decent human translation: (2) a. a". b.
Yesterday we the entire hour rolled themselves on a boat, Yesterday we went out boating for a whole hour, The final stage of landing is achieved with the aid of the fire system of high intensity, b'. The final stage of aircraft landings is accomplished with the help of high intensity lights.
(2a, b) are just not typical of human errors. It may be objected that SYSTRAN is not exactly state-of-the-art as far as MT research is con cerned, but I have deliberately chosen a system which has been extensively used in real translation situations. As a hardly original conclusion to this section, I would suggest that MT in its present situation cannot offer human-like translations, so from this point of view it does not constitute a model of the human transla tion process.
Modelling the translator’s knowledge I move on now to a consideration of (lb), noting first the observation of Johnson and Whitelock that “What we try to do when we build an MT
160
Unity in Diversity
system is to incorporate all or part of the translator’s expertise into a computer program” (1987:136). They postulate five kinds of knowledge on the part of a professional technical translator: (3) a. b. c. d. e.
target language knowledge text-type knowledge source language knowledge subject area (‘real world’) knowledge contrastive knowledge.
Let us examine these in the light of standard MT architectures: (4) a. source text ANALYSIS => IR => TRANSFER => IR SYN THESIS => target text b. source text => ANALYSIS IR => SYNTHESIS => target text In each case, IR stands for ‘intermediate representation’. (4a) shows the transfer scheme, with a bilingual component, the intermediate repre sentations being language-specific; (4b) shows the interlingual arrangement, where the IR is an interlingua intended not to be languagespecific (for m ore on these concepts, see H utchins & Som ers 1992:73-77). I shall concentrate on the two approaches seen in (4), as space does not permit consideration of recent work in MT such as shake-and-bake and co-description, or new paradigms such as examplebased MT. Let us for now consider (4) as specifying the components of the sys tem, rather than how the components relate in the translation process, which I will deal with below. If we compare (3) and (4) at the broadest level, it is plain that the translator’s source- and target-language knowl edge is captured in the analysis and synthesis components of (4). Contrastive knowledge is reflected in the transfer module of (4a), where as a strict interlingua system has no place for bilingual knowledge. Text-type and real-world knowledge can be included in MT systems, and many systems are, for instance, aimed at particular sublanguages. How ever, real-w orld know ledge is som ew hat open-ended and so is problematic. An important additional point is that human knowledge is far more flexible and less compartmentalized than that of a machine (Sager 1994:247). The more narrowly linguistic elements, viz. (3a, c and e), deserve fur ther discussion. (3a) and (3c), the monolingual parts, can be divided in the first instance into grammatical and lexical knowledge, with the former again being divided into morphological, syntactic and semantic. Not sur prisingly, these divisions are reflected in the organization of linguistic modules in MT systems, although the amount of semantic information
Bennett: MT as a M odel o f Human Translation
161
may vary. The question of contrastive knowledge - which, after all, is the one characteristic of translation - is less straightforward. Particularly in multilingual systems, one goal is to reduce the size of the transfer modules. In Eurotra (see Allegranza et al 1991), the intention was to limit transfer to simple lexical transfer (statements of the form cup => tasse). This would mean that, at best, there was no need to give statements of context in lexical transfer rules or to have rules which altered structure; to some extent I shall focus on Eurotra here, as it is the MT system I know most about. The intermediate representations were languagespecific primarily as far as lexical items were concerned (for brief motivation, see Bennett 1994:14-16). One consequence of this is that it is difficult to exploit languagepair specific translation strategies above the lexical level. So one cannot directly incorporate rules of thumb such as ‘SL structure X is regular ly translated as TL structure Y’ (e.g. a French active clause with on as subject rendered as an English agentless passive), since (a) transfer mod ules act on intermediate representations which neutralize many surface aspects of structure, and (b) they ought not to contain such structurechanging rules anyway. This conclusion would only be avoided in the case of major structural differences which were not neutralized in the IR, such as the rendition of to like as p la ire or p iacere (with switched-round arguments), or the translation into Romance languages of motion expres sions such as swam across the river (cf. (6c) below). Let’s take one more specific example of human contrastive knowl edge: In Romance language medical texts, you normally assume that a SL noun plus adjective group, e.g. radioactivité plasmatique , is going to be switched round to a noun-plus-noun compound, ‘plasma radio activity’, provided the SL adjective is formed from a noun of substance. (Newmark 1988:213) Adjectives such as p la sm a tiq u e are commonly known as relational adjectives; the correspondence noted by Newmark applies not just to medi cal language. In Eurotra research on compounds (see Bennett 1993b, Carulla 1994), such cross-linguistic differences were solved by repre senting relational adjectives as their underlying nouns in the intermediate level, e.g. plasm a for plasm atique. This permits simple lexical transfer although the problem of when to produce a relational adjective in synthe sis remains. My point here, however, is just that the piece of knowledge set out by Newmark is not captured as such, rather it is spread over sev eral modules (French analysis says such adjectives are represented as nouns, while English synthesis says modifying nouns in compounds are generally realized simply as nouns); French-English transfer does not deal
162
Unity in Diversity
with this relation. This conclusion holds a fo rtio ri for interlingual systems which have no transfer components at all. I take it, then, that one big difference be tween the knowledge of an MT system and of a human translator is that the contrastive knowledge of the latter is far richer and not purely lexical. Expert translators presumably acquire masses of instances of the kind of specific knowledge illustrated in the passage from Newmark. Of course, one might argue from this that MT systems should be bilingual only, or that only bilingual systems merit comparison with human translation. One level of linguistic knowledge omitted in the discussion above was that of text or discourse structure. If present, this could be spread over (3a, c, e) in an MT system, but generally MT systems pay little at tention to textual coherence or cohesion. I proposed (Bennett 1994) that discourse considerations are best confined to synthesis, as they are TLdependent. The fact rem ains, however, that MT systems are less successful in representing this kind of knowledge - understandably so, given the subjective nature of many of the concepts used in linguistic research on these topics. To conclude this section, it seems reasonable to say that MT sys tems are partial models of the human translator’s knowledge, in that they represent knowledge of TL and SL grammar and lexicon, plus some texttype and real-world knowledge. But, while they necessarily contain contrastive information, this is, in the ideal case, far sparser than with human translators.
Modelling the translation process Our final aspect of modelling is where a system translates in the same way a human does, going through the same processes and not just us ing the same knowledge. An immediate problem with discussing this is that relatively little seems to be known with any degree of confidence about the way people translate. The fact that translation is a conscious, deliberate process (see above) seems to make it less amenable to the psycholinguistic techniques used to study human sentence processing and production. For instance, I am not aware of any evidence concerning the time taken to translate different kinds of structure, e.g. as to whether ex amples where a phrase (italicized in (5)) is a long way from its canonical position (shown by a bullet) take a relatively long time to translate (see further below): (5) a. What do you think John is intending to do • next? b. M ary appears to be believed to have been arrested • For instance, one might wonder whether (5a) is easier/quicker to trans
Bennett: MT as a Model o f Human Translation
163
late into a language with w/z-fronting than one without, such as Japanese, and how difficult is translation of (5b) into a language like French that does not allow such repeated raisings. Such questions might be compar ed to psycholinguists’ (former) investigations into the Derivational Theory of Complexity, although this has now lost its appeal. Because of the problems of experimental study of translating, investi gators have tended to rely on think-aloud protocols whereby translators report on their mental processes while translating (e.g. Lorscher 1991). For the sake of concreteness, though, I shall here take the model offered by Bell (1991, ch. 2), which is based more on research in psycholinguistics, and less on work specifically relating to translation. It would follow from (4a) that MT systems perform translation in three sequential stages, analysis - transfer - synthesis. A naive initial view would be that human translators do the same thing, though the stages will overlap with each other rather than being sequential (Lorscher 1991:17). Bell, however, offers an analysis-synthesis picture more like (4b). The analysis procedure he sees as common to both translator and monolingual reader, though this is itself a controversial assumption, and one that needs testing. Analysis consists of syntactic, semantic and prag matic analysis, though Bell stresses that these processes are integrated, not simply consecutive. The output is a semantic representation which contains information about clause structure, propositional content, dis course structure, register and illocutionary force. This representation is now handed over to the synthesis component, whereby pragmatic, se mantic and syntactic synthesizers produce the TL text; synthesis is discussed at much shorter length, though. His sample semantic represen tation (p. 67) does contain some words of English, so I assume that he is not claiming that these representations are entirely in some universal se mantic metalanguage. In any case, I find it surprising that his model reduces a translator’s expertise to knowledge of SL and TL, with no ap parent place for contrastive knowledge. Leaving this aside, though, how can MT systems be said to measure up to this model? One large difference is that the pragmatics modules of Bell’s model have little if any place in most MT systems (see above). Another is that processing in NLP systems in general, not just in MT systems, tends to be much more sequential (e.g. first syntax, then semantics). This is particu larly so in the case of the stratificational approach of Eurotra where analysis involves assigning a surface syntactic analysis, then a relationallybased one, then the more semantic-oriented IR, with synthesis going through the reverse steps. In lacking a transfer module, Bell’s model is clearly close to that of the interlingua architecture, (4b). One particular question prompted by these remarks is the nature of synthesis in MT, in particular how it relates to synthesis in other areas of NLP, especially
164
Unity in Diversity
what kind of representation is seen as the input to synthesis. I would also argue that the idea of translating compositionally, via the combina tion of ever-larger translation units (Bennett 1994) may be common to both MT and human translation. We should however note the claim (Somers 1993) that, for human translators, structure-bound translation is the last resort, rather than the first choice. Rather than continue with such reasoning, however, I wish to pursue a slightly different line. I discussed above the idea of transfer and simple vs. complex transfer. It seems to me that a worthwhile question to ask is whether human translation has any analogue of the process of complex transfer. In other words, do cases where ST and TT differ greatly in terms of morphosyntactic structure take longer to translate than those where ST and TT are parallel in terms of surface structure, or those where they differ on the surface but are similar at the IR level (cf. the example of radioactivité plasm atiqu e )? I would suggest that these are the kinds of question that MT opens up for research into the translation process. And given that different MT systems will offer differing classifications of in dividual cases, they may help to show which architecture or system is closest to modelling the human translator. Let’s make this point more precise. ST-TT relations may be divided for these purposes into three types, bearing in mind that the notion of ‘same/different IR’ is system-dependent: (6) a.
Same syntactic structure in both: Charles works in London. / Charles travaille à Londres.
b. c.
Different syntactic structure, same IR: I believe Charles to be rich. / Je crois que Charles est riche. Different at both syntactic and IR levels: Charles ran across the street / Charles a traversé la rue en courant.
One might hypothesize that (6a-c) represent increasing difficulty for translation, and correspondingly take longer to translate. I do not know if this is correct, and the experimental problem would be to distinguish dif ficulty of translation from difficulty of analysis, but this ought in principle to be possible. One could add further cases, e.g. those where the TT con tains more information than the ST (e.g. about number or aspect). Some other examples of complex mappings worth investigating from this point of view are listed in (7): (7) a.
Simple predicate to support verb: C harles m urdered M arie. / Charles a commis un meurtre contre M arie.
Bennett: MT as a Model o f Human Translation
165
b. Verbal category expressed by inflection or via an auxiliary: M arie w ill leave soon. / M arie p a rtira bientot.
Some of the restructurings examined in Tsutsumi (1990) are also rel evant here. MT-oriented linguistics, therefore, can offer a classification of constructions which may be tested for their processing difficulty. Without far more being known about how people translate, it is hard to discuss whether MT systems actually model the way human transla tors work. I therefore see this research programme as an instance of “simulation of poorly understood systems” (Simon 1981): this can still be useful, especially if we are willing to abstract from the detail of some set of phenomena, which makes modelling easier.
Concluding remarks So does MT provide a model of human translation? I hope it will be clear that no simple yes or no answer is appropriate. MT systems do not model human translators in the sense of producing human-standard trans lations or making human-type errors. They may contain the same kind of information/knowledge possessed by human experts, but they often divide this up differently. It is hard to say whether MT systems translate the same way people do, but one useful question raised by our study is whether there is a human equivalent of complex transfer, and whether what takes a person longer to translate also takes a machine longer. At the very least, MT provides for the formulation of questions that the empirical study of human translation should address.
Acknowledgements I am very grateful to Juan Sager and Harry Somers for comments on an earlier version.
References Allegranza, V., P. Bennett, J. Durand, F. Van Eynde, L. Humphreys, P. Schmidt and E. Steiner (1991) ‘Linguistics for Machine Translation: the Eurotra Linguistic Specifications’, in C. Copeland, J. Durand, S. Krauwer and B. Maegaard (eds) The Eurotra Linguistic Specifications , Luxembourg: Office for Official Publications of the European Communities, 15-123. Bell, R. (1991) Translation and Translating, London: Longman. Bennett, P. (1993a) ‘A Comparative Evaluation of the Eurotra Preference Mechanism’, in P. Bennett and P. Paggio (eds) Preference in Eurotra , Luxembourg: Office for Official Publications of the European Commu nities, 91-103.
166
Unity in Diversity
Bennett, P. (1993b) ‘A Multilingual Translation-oriented Typology of Com pound Nouns’, Traitement Automatique des Langues 34(2):43-58. Bennett, P. (1994) The Translation Unit in Human and Machine\ Babel 40:1220 . Carulla, M. (1994) ‘Relational Adjectives in the Translation from Germanic Nominal Compounds into Romance Languages’, in Proceedings o f the Workshop on Compound Nouns: Multilingual Aspects o f Nominal Com position, Geneva: IS SCO. 103-107. Fodor, J. (1983) The Modularity o f Mind, Cambridge, MA.: MIT Press. Hutchins, W. J. and H. L. Somers (1992) An Introduction to Machine Trans lation, London: Academic Press.
Johnson, R. (1983) ‘Parsing - an MT Perspective’, in K. Sparck Jones and Y. Wilks (eds) Automatic Natural Language Parsing, Chichester: Ellis Horwood, 32-38. Johnson, R. and P. Whitelock (1987) ‘Machine Translation as an Expert Task’, in S. Nirenburg (ed) Machine Translation: Theoretical and Methodologi cal Issues, Cambridge: Cambridge University Press, 136-144. Knowles, F. (1979) ‘Error Analysis of SYSTRAN Output’, in B. Snell (ed) Translating and the Computer, Amsterdam: North-Holland, 109-133. Lörscher, W. (1991) Translation Performance, Translation Process, and Translation Strategies, Tübingen: Gunter Narr. Marcus, M. (1980) A Theory o f Syntactic Recognition fo r Natural Language, Cambridge, MA.: MIT Press. Newmark, P. (1988) A Textbook o f Translation, London: Prentice-Hall. Sager, J. C. (1994) Language Engineering and Translation, Amsterdam/Phila delphia: John Benjamins. Simon, H. (1981) The Sciences o f the Artificial, Cambridge, MA.: MIT Press, 2nd edition. Somers, H. L. (1993) ‘Current Research in Machine Translation’, Machine Translation 7: 231-246. Thompson, H. (1983) ‘Natural Language Processing: a Critical Analysis of the Structure of the Field, with Some Implications for Parsing’, in K. Sparck Jones and Y. Wilks (eds) Automatic Natural Language Parsing, Chichester: Ellis Horwood, 22-31. Tsutsumi, T. (1990) ‘Wide-range Restructuring of Intermediate Representa tions in Machine Translation’, Computational Linguistics 16:71-78.
Section Five Interpreting Studies
Page Intentionally Left Blank
Unity in Diversity The Case of Interpreting Studies FRANZ POCHHACKER Department o f Translation and Interpreting University o f Vienna, Austria Research in interpreting, which dates back to the 1950s, has gained renewed momentum since the late 1980s. The article examines this ‘Renaissance p e rio d ’ o f interpreting studies under the theme o f ‘Unity in Diversity ’, discussing the conceptual and methodological common ground within this emerging (sub)discipline o f academic study. Analysis o f the recent literature shows that there is a broad consensus within interpreting studies as to what is to be studied. There is less agreement on how that object o f study ought to be approached. Following a discussion o f different research paradigms and their relative status within interpreting studies , one possible way o f overcoming the ‘battle o f the paradigm s' in (conference) interpreting research is suggested. The article argues that, given the intrinsic diversity o f its object o f study, the interpreting research community should not push fo r greater uniformity o f methodologi cal approach but instead turn its diversity into a strength by discovering new relationships and links, thus reinforcing the com munity ’s sense o f unity and internal coherence.
Introduction Research efforts aimed at analyzing and explaining the phenomenon of interpreting, i.e. oral translation, date back to the 1950s. The time course of interpreting research as an area of scholarly study has thus been paral lel to that of the systematic investigation of written translation. In terms of its conceptual focus and methodological orientation, however, the study of interpreting has developed along strikingly different lines. Several broad historical categories or periods of interpreting studies have been suggest ed by Gile (1994), who views the evolution of ‘Interpretation Research and Theory’ in the late 1980s and early 1990s as a period of ‘Renais sance’ and ‘renewal’ (Gile 1995). The present article is an attempt to examine this ‘Renaissance period’ of interpreting studies under the theme of ‘Unity in Diversity’ in order to assess the coherence, or lack thereof, of interpreting studies as an emerg ing (sub)discipline of academic study. In other words: to what extent is there enough conceptual and methodological common ground in inter preting studies to speak of a “disciplinary utopia” in the sense of “a shared
170
Unity in Diversity
interest in a common set of problems, approaches, and objectives” (Holmes 1988:67)? In trying to answer this question, I shall discuss the theoretical and methodological approaches to the study of interpreting as well as the object of study itself, on the basis of bibliographical analyses and se lected metatheoretical writings. It will be my paradoxical contention that one way of strengthening the coherence within the largely disjointed in terpreting research community may be to develop a heightened sense of awareness - and acceptance - of the intrinsic diversity of the phenom enon of interpreting and, thus, the discipline of interpreting studies.
(Conference) interpreting studies In a bibliographical analysis of recent (1989-1994) contributions to the literature on interpreting (Pôchhacker 1995a) a break-down of some 400 specific titles on interpreting by mode and type of interpreting (si multaneous vs. consecutive; conference, court, bilateral, sign-language, media, at-sight) yielded a clear predominance of references dealing with conference interpreting. The roughly 200 works on simultaneous and consecutive interpreting, which were found in the corpus in a ratio of 3:1, together with the 83 references focusing explicitly on interpreting in the conference setting, account for more than two thirds of the 414 themati cally specific items. The remaining entries pertain to court interpreting (50), bilateral interpreting (32), sign-language interpreting (12), media interpreting (10) and interpreting at sight or sight translation (8). While publications on simultaneous interpreting can conceivably refer also to work in the courts or in non-conference settings where whispered simultaneous is used, there is little evidence of such cases in the literature. It may therefore be fair to say that the recent literature on interpreting reflects a high degree of uniformity and common ground with regard to the mode and type of interpreting under study. At least from a European perspective, few would assume otherwise. After all it was the striking phenomenon of simultaneous interpreting which, some 30 years after initial trials with simultaneity in the late 1920s, generated the first wave of systematic interpreting research, carried out by psychologists and psycholinguists. As the simultaneous mode quickly proved more efficient in conference-type settings, newly founded schools for interpreter training, many of which came to be established at univer sity level right after or even during World War II, could hardly afford to ignore simultaneous interpreting in their curricula. Not surprisingly, train ing in ‘the booth’ has been a central concern not only in the curricula but also in the research activities of interpreter training institutions such as those affiliated with C.I.U.T.I. ( Conférence internationale perm anente d ’instituts universitaires de Traducteurs et Interprètes) with which the
Pochhacker: Unity in Diversity in Interpreting Studies
171
most productive figures in interpreting research are associated (Pochhacker 1995b:49f). For these and other reasons, the interpreting research community, as reflected in its body of recent literature, boasts a far-reaching consensus as to what is to be researched: the basic object of study is (simultaneous and consecutive) conference interpreting. While there is an astonishingly broad range of specific research topics, there is also wide-spread agree ment on the (half a dozen or so) fundamental questions to be asked and answered (cf. Shlesinger 1995:8). Where the (conference) interpreting research community treads on less common ground is the theoretical and methodological approach(es) to be taken. The argument over how the object of study ought to be approached has lately been framed in terms of a dichotomy between ‘scientific research’ and ‘speculative theorizing’.
The solid soil of science Starting with a paper by Gile (1990), which pits ‘personal theorizing’, as represented by Seleskovitch and followers, against empirical work by the basic rules of scientific investigation, it has become common to view the interpreting research community in the early 1990s as roughly di vided into two camps: the first group prefers explorations which require precision of logical processes, and where members are interested in the natural sciences and quantification; the second group prefers explora tions which involve the intellect in a less logically rigorous manner, where members are interested more in a liberal arts approach and general theo rizing (Moser-Mercer 1994:17). Moser-Mercer (1994) views the two (sub)communities in terms of distinct paradigms, hinting, perhaps, not only at their apparent incompat ibility but also at their competitive struggle in the Kuhnian process of scientific (r)evolution. This ‘battle of the paradigms’ could also be de duced from Gile (1995), who sees the liberal arts community in a “process of attrition”, “losing ground”, losing power and “losing influence over the interpreting research community” (cf. Gile 1995:16ff, 25f). He seems to regard “liberal-arts-type essays” as a species of publications one would like to see extinct and stresses the high appreciation of “contributions from other disciplines, in particular cognitive psychology and psy cholinguistics” (Gile 1995:17f). If one adds to this the explicit preference for “knowledge achieved through experimental research” (Lambert 1994:6) which underlies a recent volume on em pirical interpreting re search, one could well get the impression that interpreting studies is moving toward the “winning paradigm” of “scientific experimental meth ods” (Fabbro and Gran 1994:307). This impression may actually be no more than that - an impression
172
Unity in Diversity
possibly derived from an all too selective reading of the recent meta-theoretical literature. Still, the impression is apparently strong enough within the interpreting research community to draw an outspoken critical response such as that voiced by Garcia-Landa (1995). Considering interpreting a “social process” rather than a “natural object”, Garcia-Landa (1995:395f) argues that the interpreting research community cannot do justice to its object of study by resorting to the epistemological model of the natural sciences. With his attack on the “naturalistic credo” Garcia-Landa (1995:392) could be said to “reject the idea of being strait-jacketed by one single intellectual or methodological approach”, were it not for the fact that the latter quotation is actually borrowed from the supposedly opposite camp: it is taken from a similarly outspoken critical response by Gran and Snelling (1993) to a paper calling for a more “comprehensive analytical perspective” or “theoretical framework” and a “prior theoreti cal model or conceptualization” in empirical research on interpreting (Pochhacker 1993:52, 55). Is this yet another example of the two camps, the liberal arts scholars and the natural science group, telling each other off across an unbridge able methodological gap? Speaking for myself, I can safely say that the intention in Pochhacker (1993) was clearly not to reject or downgrade empirical research efforts and preach the gospel of the ‘omni-comprehensive theory’ - though I have to concede that my liberal use of the word theory in assertions about (the lack of) ‘a common theoretical frame w ork’ proved regrettably m isleading. Com pared to that less than successful attempt of mine to call for some generally agreed conceptual and methodological ‘terms of reference’, Garcia-Landa’s (1995) appeal to both the theorizing and the research kind within interpreting studies is perfectly clear: “the young research community we are talking about would do well to stop squabbling about who is doing the right thing and agree on a minimum of rules to start thinking and research the object of translation theory” (Garcia-Landa 1995:394). Against this backdrop, I would like to suggest that the apparent con troversy within the (conference) interpreting research community about who has the right approach and which is the sufficiently scientific meth odology could perhaps be defused by broadening the conceptual terms of reference for interpreting studies in general, i.e. by working towards an increased awareness and acceptance of the intrinsic diversity of inter preting and thus of the diverse methodological approaches appropriate for its investigation.
There’s more to interpreting ... In a handbook on careers in world affairs Moser-Mercer (1993:96) de
Pôchhacker: Unity in Diversity in Interpreting Studies
173
fines interpreting as “the oral rendition of statements spoken in one lan guage into another language”. This working definition in itself would not raise any eyebrows; its continuation, however, may seem rather striking, at least from the classic conference interpreting perspective typical of CIUTI schools: “This can be carried out in a variety of settings: a hospi tal, a court of law, a business meeting, an international conference, on TV, or even over the telephone where, by means of a conference call, two parties communicate with the help of an interpreter”. If we agree that interpreters in all the above settings are essentially doing the same thing, i.e. interpreting, do we also accept the corollary that interpreting studies is charged with accounting for all facets of its highly diverse object of study? Is the interpreting research community, now largely a conference interpreting research community, an ‘open society’, able and willing to accommodate ‘immigrants’ on their own terms? And to what extent do the latter assimilate the body of experience and knowl edge of the more established community? One could muse at length on the metaphors used to describe the dif ferent approaches to interpreting studies: does the discipline represent a monocultural melting-pot or a multi-ethnic salad-bowl? Lest this seem like taking the liberal-arts-type essay too far, I would like to put the saladbowl hypothesis to an empirical test by analyzing three contemporary monographs on ‘off-conference’ interpreting for the bibliographical com mon ground they share with the mainstream literature on interpreting as well as with one another. In 1992, three English-language monographs - two Ph.D. disserta tions and one textbook - on interpreting were published: Cokely (1992) on sign-language interpreting, de Jongh (1992) on court interpreting, and Wadensjô (1992) on dialogue interpreting (community interpreting). As one might expect, the two Ph.D. dissertations each contain a section de voted to a review of research on interpreting (Cokely 1992:5-17; Wadensjô 1992:17-23) while the introductory textbook on court interpreting has a more practical orientation, with more than half of its roughly 300 content pages devoted to exercise material and useful documentation. Even so, the number of references listed in the bibliography is higher than in the two dissertations (cf. the bibliographical data summarized in Table 1). What is of interest here is the extent to which these books draw on the ‘mainstream literature’ on interpreting. The reference corpus chosen for this analysis is the B ibliography on Interpretation as published in issue no. 1 (1988) of The Interpreters' N ew sletter. The Bibliography, which was essentially compiled in the 1980s by a team headed by Jennifer Mack intosh for the A ssociation Internationale des Interprètes de Conférence (AIIC, the international association of conference interpreters), contains 318 entries for articles, books, reviews and theses on interpreting. (For a
174
Unity in Diversity
more detailed description and analysis of this bibliographical corpus, see Pochhacker 1995a.) Table 1: Bibliographical data in three monographs on interpreting
N o . o f p a g e s:
de J o n g h
W a d e n sjo
C o k e ly
(1 9 9 2 )
(1 9 9 2 )
(1 9 9 2
3 3 8 (+ in d e x )
285
199
N o. o f
2 1 7 ( w /o n e w s
17 0
172
r e fe r e n c e s :
p a p ers and f ilm s )
N o. o f N on -
21
29
9
E n g lis h
13 F ren ch
12 S w e d is h
4 F ren ch
r e fe r e n c e s :
N o . o f E n g lis h
5 S p a n is h
8 R u s s ia n
3 G erm an
2 R u s s ia n
8 G erm a n
2 D u tc h
1 G erm a n
1 N o r w e g ia n
196
141
163
r e fe r e n c e s :
Table 2: Overlap between bibliographical references in three monographs on in terpreting and the ‘Bibliography on Interpretation ’
'Bibliography
de Jongh
Wadensjo
Cokely
(1992)
(1992)
(1992)
Titles:
36 (11.3%)
2 (0.6%)
16 (5%)
Authors:
32 (21.3%)
4 (2.7%)
16 (10.7%)
on Interpretation'
de Jongh
6 titles
21 titles
(1992)
9 authors
23 authors
Wadensjo (1992)
5 titles 14 authors
The first section of Table 2 shows the number (and percentage) of titles and authors in the respective monograph which are also listed in the B ibliography on Interpretation : only two of Wadensjo’s and sixteen of Cokely’s references are drawn from the body of literature brought to
Pochhacker: Unity in Diversity in Interpreting Studies
175
gether in the 1988 Bibliography; the picture for the authors cited is rough ly similar. The second part of Table 2 is a matrix for matching titles and authors between two monographs each. De Jongh and Wadensjo have six titles and nine authors in common; Cokely and Wadensjo share five titles and fourteen authors. The two books written in the U.S., Cokely and de Jongh, contain twenty-one identical references. Since a more detailed discussion is beyond the scope of this article, these quantitative indicators will have to suffice to support the basic ar gument that the range of theoretical and methodological approaches to interpreting research - and the range of bibliographical (re)sources proves refreshingly broad and diverse once we transcend the established paradigms of (conference) interpreting research and focus on the activity of interpreting in different institutional settings and socio-communicative circumstances.
Interpreting research commUNITY in diversity? Deliberately widening the focus of interpreting studies to include offconference forms of interpreting, which tend to be linked more closely to social, legal, ethical and cultural issues than to models of cognitive pro cessing, could be a way for our emerging discipline to turn the liability of diversity into a unifying strength: the more the previously neglected research issues are brought to the fore, the sooner there will be a full consensus that - a variety of (cf. Shlesinger 1995) - experimental cognitive-science approaches are very well suited to the study of some aspects of interpreting just as a variety of investigations informed by insights and models from translation theory, communication studies, discourse analy sis, semiotics, sociology, anthropology, etc., can be most appropriate for other facets of our object of study. The crucial problem for such an interpreting research commUNITY in diversity will no doubt be that of speaking the same language. Within the simultaneous conference interpreting research community alone, the intricacies of specific terminologies and methodologies can be bewilder ing even to the interested reader (cf. Shlesinger 1995). Would it be too optimistic to hope that the interpreting crowd, more than anyone, should be capable of rising to this challenge? From the point of view of the emerging discipline, opening up within and across interpreting studies would seem worth the effort. While pushing for greater uniformity of methodological approach could come at the cost of neglecting or even excluding relevant phenomena, the attempt to come to terms with the full diversity of its object of study could prompt the interpreting research community to discover new relationships, establish additional links, and thus strengthen its sense of unity and internal coherence.
176
Unity in Diversity
References Cokely, D. (1992) Interpretation: A Sociolinguistic Model, Burtonsville MD: Linstok Press. de Jongh, E. M. (1992) An Introduction to Court Interpreting. Theory and Practice , Lanham, New York & London: University Press of America. Fabbro, F. and L. Gran (1994) ‘Neurological and Neuropsychological As pects of Polyglossia and Simultaneous Interpretation’, in S. Lambert and B. Moser-Mercer (eds) Bridging the Gap. Empirical Research in Simul taneous Interpretation , Amsterdam/Philadelphia: John Benjamins, 273-317. Garcia-Landa, M. (1995) ‘Notes on the Epistemology of Translation Theory’, Meta XL(3):388-405. Gile, D. (1990) ‘Scientific Research vs. Personal Theories in the Investiga tion of Interpretation’, in L. Gran and C. Taylor (eds) Aspects o f Applied and E xperim ental R esearch on Conference In terpretation , Udine: Campanotto, 28-41. Gile, D. (1994) ‘Opening up in Interpretation Studies’, in M. Snell-Homby, F. Pöchhacker and K. Kaindl (eds) Translation Studies. An Interdiscipline , Amsterdam/Philadelphia: John Benjamins, 149-158. Gile, D. (1995) ‘Interpretation Research: A New Impetus?’, Hermes. Journal o f Linguistics 14:15-29. Gran, L. and D. Snelling (1993) ‘Editorial’, The Interpreters’s Newsletter 5:1-2. Holmes, J. S. (1988) Translated! Papers on Literary Translation and Trans lation Studies , Amsterdam: Rodopi. Lambert, S. (1994) ‘Foreword’, in S. Lambert and B. Moser-Mercer (eds) Bridging the Gap. Empirical Research in Simultaneous Interpretation , Amsterdam/Philadelphia: John Benjamins, 5-14. Moser-Mercer, B. (1993) ‘Translation, Interpretation and Terminology. In troduction’, in P. Gerard (ed) Guide to C areers in W orld A ffairs , Washington: Foreign Policy Association, 95-101, 3rd edition. Moser-Mercer, B. (1994) ‘Paradigms Gained or the Art of Productive Disa greement’, in S. Lambert and B. Moser-Mercer (eds) Bridging the Gap. Empirical Research in Simultaneous Interpretation , Amsterdam/Phila delphia: John Benjamins, 17-23. Pöchhacker, F. (1993) ‘On the “Science” of Interpretation’, The Interpreters’ Newsletter 5:52-59. Pöchhacker, F. (1995a) ‘Writings and Research on Interpreting: A Biblio graphic Analysis’, The Interpreters’ Newsletter 6:17-32. Pöchhacker, F. (1995b) “‘Those who do...”: A Profile of Research(ers) in Interpreting’, Target 7(l):47-64. Shlesinger, M. (1995) ‘Stranger in Paradigms: What Lies Ahead for Simulta neous Interpreting Research?’, Target 7(l):7-28. Wadensjö, C. (1992) Interpreting as Interaction. On Dialogue-Interpreting in Immigration Hearings and Medical Encounters, Linköping: Linköping University.
Language Direction and Source Text Complexity Effects on Trainee Performance in Simultaneous Interpreting JORMA TOMMOLA AND MARKETTA HELEVÀ Centre fo r Translation and Interpreting University o f Turku, Finland The general consensus among professional simultaneous interpret ers and those who run training programmes is that interpretation should be into one’s mother tongue or A language. Others argue that the efficiency o f creating a rich semantic representation o f the source text is the central aspect o f interpretation and it is reason able to suppose that the information density or conceptual complexity o f the source text will have an effect on interpreter performance , irrespective o f language direction. This article describes a study that measured the effects o f language direction (English-Finnish or Finnish-English) and source text complexity on the performance o f 12 trainee interpreters. Complexity produced a sig-nificant effect on performance. In the case o f language direction , interpreting from A t o B appeared in certain instances to produce a more satisfactory result. The article goes on to deal with issues relating to the de pendent variable , sample size and comparison o f trainees and professionals.
Introduction In this article we report a small-scale study on the performance of trainee simultaneous interpreters, where we examined the effects of two independ ent variables. The first of these, language direction, is a task variable, and involves two conditions, interpreting either from one’s non-dominant (B) into the dominant (A) language, or from A into B. The second variable is the complexity of the source text (ST) to be interpreted. The notion of text complexity has often been seen in terms of either syntactic or seman tic difficulty. In this article we will not deal with syntactic and semantic difficulty separately, and will accept that the basically syntactic modifi cations we introduce in the experimental materials will also have some semantic consequences for the comprehension and interpretation pro cesses. We will employ the term linguistic complexity to denote this combined effect of syntax and meaning. With respect to our first independent variable, language direction in interpreting, the general opinion among interpreters, their professional organizations, different training programmes, and employers such as the
178
Unity in Diversity
various institutions of the European Union normally favours interpreta tion into one’s mother tongue (or the dominant A language). The familiar argument is that, in the non-dominant language, the level of receptive skills and knowledge will always be higher than the level of productive skills. Accurate comprehension in the non-dominant B language will be achieved with a greater likelihood than its accurate production, and it is thus advantageous to have the B-language text as the source text to be understood, and have the message formulation stage take place in the dominant language (cf. Gervern 1976). The resulting target text will thus be more likely to contain both the necessary information and the stylistic nuances needed to transmit the source-text speaker’s communicative intentions. On the other hand, however, one might also want to emphasize the advantages of the speedy and fully accurate comprehension of the mother tongue. Research on bilingual language processing (e.g. Dornic 1978, and many others) indicates that the level of automatization in mothertongue comprehension is higher than that of the less dominant language, even though communicatively the bilingual might seem to handle the two languages equally successfully. Particularly if the text to be interpreted is technical rather than general in content, it is possible that the A to B di rection can produce a more satisfactory result. Retour interpretation (A to B) is sometimes also a practical necessity, as in the case of the Finnish language in the European Union. When interpreters with adequate B knowledge of a small language are not available, interpreters with this language as their A language will have to work in both directions. The question of language direction in interpreting thus has practical consequences for the profession and work opportunities of an interpreter, and also has some theoretical interest. Empirical research on this ques tion, however, is not particularly abundant, and evidence often tends to come from incidental comments made by researchers concentrating on error analyses or the investigation of temporal or communicative aspects of interpreting, or from the pedagogical convictions of trainers. Further exploration of this question thus seems warranted. In the following we present a set of data obtained from trainees whose B language (English), while fluent, is still non-dominant with respect to their mother tongue, Finnish. In the analysis of the performance of the subjects, the emphasis will be on the amount of content information conveyed. The stylistic ef ficiency of the delivery, or the requirement of linguistically faultless output, although important in practice, are not the main focus of this ex periment. If comprehension is central for the transmission of informational content, one might expect that, for trainees, going from A to B might result in a more accurate performance than going in the standard direc tion of B to A.
Torninola & Helevà: Language Direction & ST Complexity
179
Complexity, our second independent variable, was defined above as the combined effect of sentence structure and meaning. Numerous psycholinguistic studies have attempted to demonstrate the effects of these two sources on comprehension. The typical and expected result with re spect to the syntactic structure of sentences is that increasing the complexity of a paragraph by combining surface clauses and sentences into one long and complex sentence increases processing difficulty in relation to a paragraph where more or less the same propositional content has been expressed in more easily manageable surface constructions (e.g. Kintsch 1974). Similarly, if the number of propositional semantic units in a paragraph is increased while holding paragraph length roughly con stant, the material takes longer to understand, and the increase in processing time per additional proposition is larger the more complex the paragraph is. Thus, the more complex the meaning representation that the comprehender has already built, the more time it takes to integrate new information into this representation. The load on the limited-capacity working memory during simulta neous interpretation is considerable, because the task involves the parallel execution of comprehension, production and monitoring processes. In addition, both SL input and TL output processes take place in the same modality, the auditory channel, which is another source of mental load. It can be expected that during such a demanding task, and particularly in the case of trainees, the complexity of the source text will have a crucial effect on the accuracy of the performance. The demand from the complex processing task alone is likely to force the trainee to operate at the limits of working memory capacity, and if load is further increased by increas ing the linguistic complexity of source text formulation, this should be reflected in the performance as a decrease in the accuracy of the output.
Method Subjects
Twelve volunteer subjects participated. All were students at the Depart ment of English Translation Studies at the Centre for Translation and Interpreting, University of Turku, and had either completed or were about to finish the 15-credit interpretation course module offered by the Department. Some of the subjects also had limited interpreting experi ence outside the study programme. All had Finnish as their mother tongue. D esign
The experiment had a two-factor within-subjects design. The two inde pendent variables were Linguistic Com plexity (sem antically and
180
Unity in Diversity
syntactically Simple or Complex), and Language Direction (EnglishFinnish or Finnish-English). The dependent variable was a Propositional Accuracy Score, i.e. the proportion of source text propositions (cf. sec tion M aterials below) correctly rendered in the interpretation. M aterials
Four texts dealing with international economics were modified from re views published by two major Finnish banks to resemble conference speeches. The texts were relatively general and did not include jargon specific to economics. Nevertheless, a word list of key terms was given to the subjects one week before the experiment. English-language versions of the texts were available in the translated versions of the same publications ( Unitas: Finnish Economic Q uarterly R eview and K O P Economic Review). The translations were also modi fied for oral presentation, and an attempt was made to achieve maximally close word-for-word correspondence with the Finnish original. The av erage duration of the texts was approximately 4 minutes. Starting from the second paragraph of each text, the complexity of the beginning of each further paragraph was systematically varied in the fol lowing way. In the Simple version, the paragraphs began with a sentence triplet consisting of either simple sentences or main clauses with only one subordinate clause. In the Complex version, the information corre sponding to the triplet was expressed within a single sentence, involving chiefly relative clauses and other embedded clauses, and nominalisations. The result was a decrease in redundancy, and a more convoluted surface expression of the propositional content. In some cases, the variation of complexity necessarily also introduced slight shifts in thematics in addi tion to the purely syntactic modification, as in the following illustration from Text 3, which dealt with the problems of the Finnish textile indus try: (table No. 1) Table 1: Simple and complex versions o f a test paragraph
Simple The emergence of new producer coun tries has resulted in a change in the competition set-up. Another significant factor that has also affected this change has been the removal of import restric tions and tariff protection. The change in the competition set-up has remark ably influenced the development of the domestic textile industry.
Complex The change in the competition set-up, whose cause has been the emergence of new producer countries in addition to another significant factor affecting it - the removal of import restrictions and tariff protection - has remarkably in fluenced the development of the domestic textile industry.
Tommola & Helevà: Language Direction & ST Complexity
181
Four sets of materials were constructed, each of which had one easy Finnish, one difficult Finnish, one easy English, and one difficult English text. In all sets, the four texts appeared in the same order, but the order of the language of presentation and the order of the simple and complex versions was systematically varied to prevent practice or fatigue effects. Propositional descriptions of the elements of meaning in the target sentences were prepared using a system modified from Bovair and Kieras (1985). Earlier and on-going experiments have indicated that the propositionalization technique is relatively straightforward and reliable (cf. Tommola and Lindholm 1995; Tommola 1995). The method involves the construction of a semi-formal representation of the semantic elements in the source discourse in terms of predicates and arguments. This set of meaning elements (a text base) can then be used as a basis for evaluating the product of comprehension, recall, or interpretation processes. We will not discuss the method in detail, but Table 2 describes one way of propositionalizing the set of target sentences given above (asterisks indi cate sentence boundaries in the Simple version): Table 2: Propositional description o f simple text in Table 1
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17
EMERGE PRODUCER-COUNTRY MOD NEW PRODUCER-COUNTRY CAUSE 01 04 CHANGE SET-UP MOD COMPETITION SET-UP AFFECT FACTOR CHANGE(02) MOD SIGNIFICANT FACTOR MOD ANOTHER FACTOR REF FACTOR 10,12 REMOVE $ RESTRICTION MOD IMPORT RESTRICTION REMOVE $ PROTECTION MOD TARIFF PROTECTION INFLUENCE CHANGE(02) DEVELOPMENT MANNER-OF 14 REMARKABLE OF DEVELOPMENT TEXTILE-INDUSTRY MOD DOMESTIC TEXTILE-INDUSTRY
Although the meaning elements have here been given in English, this does not, of course, imply a hypothesis about the nature of the base list representation constructed by the interpreter when understanding the passage; the base list could well be represented in some languageindependent format. In practice, there was no difficulty in assessing the Finnish-language outputs in terms of semantic elements expressed in the above fashion.
Unity in Diversity
182
The Simple and Complex versions of the four Finnish-language texts and their close English translations were recorded by native speakers at a normal speech rate, and the speakers were asked to simulate a confer ence presentation in their delivery. From the master tape, 12 individual cassettes were prepared, in which the four texts appeared in the same order but the order of language and complexity was systematically varied. P rocedure
Subjects were tested individually in a language laboratory. Written in structions were given in Finnish before the experiment. The subject’s output was recorded on the second track of the cassette. A short pause separated the texts. The entire session lasted approximately one hour. R esults
The subjects’ performances were transcribed word for word, after which the target-sentence renderings were propositionalised and scored against the source-text list of propositions. A relatively strict set of criteria was employed; basically, credit was given to the output only if the original proposition was explicitly present. However, since simultaneous inter preting often involves summarization strategies which are employed to reduce momentary information overload, credit was also given to para phrase-type renderings which, however, unambiguously reproduced the content of the original proposition. Only one judge performed the scor ing, since previous research had found that inter-judge reliability in this scoring task tends to be high (.98 in the experiment by Tommola and Lindholm 1995). Tables 3 and 4 give the per cent accuracy scores for the two main effects. Table 3: Per cent propositional accuracy in texts with two levels o f linguistic complexity
mean
sd
Simple
52.01
11.52
Complex
42.00
14.45
Table 4: Per cent propositional accuracy with E-F and F-E directions o f interpretation
mean
sd
English-Finnish 45.38
14.59
Finnish-English 48.63
11.38
Tommola & Helevd: Language Direction & ST Complexity
183
In a two-way analysis of variance, a significant main effect emerged for Linguistic Complexity. Across Language Direction, interpretation performance with linguistically simpler source material was more suc cessful than with linguistically complex material (F = 15.96; df = 1.11; p = .002). Across the two degrees of Complexity, Language Direction failed to produce a significant effect (p = .10), although a slight trend is apparent in the means to the effect that going from A to B might produce a slightly more accurate result in this group of trainees. Figure 1 describes the interplay of the variables of Linguistic Complex ity and Language Direction in this material. When texts are linguistically simple, the per cent accuracy score is identical in both directions of interpretation, but with more complex material, there is a hint of the A to B direction being more easily manageable for trainee interpreters. The statistical interaction between Linguistic Complexity and Language Direction is, however, non-significant in this small set of data (p > .10). Figure 1: Linguistic Complexity p lo tted against Language Direction
Legend: S: linguistically simple C: linguistically complex FE: from Finnish into English EF: from English into Finnish
Discussion The results of this experiment showed that, with trainee simultaneous
184
Unity in Diversity
interpreters, linguistically more complex source texts produced a lower propositional accuracy score than did linguistically simpler and more re dundant texts. However, there was no statistically significant effect of the language direction in which interpretation proceeded, although the data revealed a slight trend suggesting that when the subjects were inter preting from their mother tongue into their B language, more propositions were correctly rendered. Furthermore, the interaction between the two variables failed to reach statistical significance. It seems plausible that conceptual and linguistic complexity of the source text is a psycholinguistic reality that affects the process of SI. When the conceptual content of the source text is complex, or when the content has been formulated in complex surface structures, the trainee interpreter may be faced with information overload which has an adverse effect on the quality of the performance. It is possible that the effect of complexity is less pronounced with experienced interpreters whose lowerlevel comprehension processes, such as lexical access or syntactic parsing, are more automatized and modularized, and whose capacity-management strategies are generally at a higher level. Further research on the effects of linguistic complexity with larger groups of trainees and profes sionals might be warranted. A statistically significant effect of language direction could not be found in this experiment. Part of the explanation for this may lie in the small number of subjects and in the relatively large standard deviations of the accuracy scores. However, the slight advantage of the A to B di rection suggests again that when the interpreter can fully and easily comprehend the source text, losses of crucial information and occur rences of misinterpretations may be rarer than in the B to A case. This may apply especially to non-professional interpreters. Again, a potential ly interesting continuation to the present study might be the comparison of sufficiently large groups of experts and novices with respect to their performance in the two language directions. The experiment also attempted to test the usability of the proposi tional accuracy score in investigations of simultaneous interpreting. In general, the technique seemed to be suitable for analyses of the process ing of source-text semantic content. However, the analysis technique disregards the hierarchical nature of the semantic representation built from the source text by the interpreter. The successful interpreter copes with memory and time limitations by using condensation and summa rization techniques (cf. Sunnari 1996), and by resorting to intentional omissions of redundant data. Since the clear transmission of gist infor mation is an essential quality of successful simultaneous interpreting, analyses of propositional accuracy should probably take it into account by giving different ratings to superordinate and subordinate propositions
Tommola & Helevà: Language Direction & ST Complexity
185
present in the output. Superordinate propositions that contain the essen tial information in an utterance could be awarded a higher score than subordinate propositions that merely add details to the basic representa tion. The exact details of a scoring system based on these considerations remain to be formulated and experimentally tested. Another inherent limitation of the method should be mentioned: the propositional accuracy score focuses merely on the amount of infor mation correctly rendered in the target language. A high propositional accuracy score does not necessarily mean that the message is conveyed in a clear and unambiguous way. Even if most of the semantic elements of the source text are present in the target message, other factors such as grammatical errors, false starts, hesitant delivery, etc., can greatly reduce the quality perceived by the listeners, and in some cases prevent accurate comprehension of the interpretation. Propositional analysis offers a promising method for analyzing the ideational or informational content of the interpreter’s performance, but it cannot be used to assess the com municative aspects of simultaneous interpretation. To achieve this, other methods, such as evaluations of the listeners’ subjective judgements of informativeness and the quality of delivery should be employed.
Acknowledgements We thank Erkki Satopàà, Rosemary McKenzie, Heikki Aaltonen and Jukka Hyonà for technical and other assistance.
References Bovair, S. and D. E. Kieras (1985) ‘A Guide to Propositional Analysis for Research on Technical Prose, in B. K. Britton and J. B. Black (eds) Un derstanding Expository Text, Hillsdale, N.J.: Lawrence Erlbaum, 315-362. Dornic, S. (1978) T he Bilingual’s Performance: Language Dominance, Stress, and Individual Differences’, in D. Gerver and H. W. Sinaiko (eds) Language Interpretation and Communication , New York: Academic Press, 259-271. Gerver, D. (1976) ‘Empirical Studies on Simultaneous Interpretation: A Review and a Model’, In R. W. Brislin (ed) Translation: Applications and Research , New York: Gardner Press, 165-207. Kintsch, , W. (1974) The Representation o f Meaning in Memory , Hillsdale, N.J.: Lawrence Erlbaum. Sunnari, M. (1996) ‘Comparison of Expert and Novice Performance in Si multaneous Interpreting’. Paper presented at the XIV World congress of the Fédération Internationale des Traducteurs, 12-16 February 1996, Melbourne. Tommola, J. (1995) ‘Gist Recall as an Aptitude Test in Interpreter Training’,
186
Unity in Diversity
in P. W. Krawutschke (ed) Connections. Proceedings o f the 38th Annual Conference o f the American Translators A ssociation , Medford, N.J.: Information Today Inc., 471-481. Tommola, J. and J. Lindholm (1995) ‘Experimental Research on Interpret ing: Which Dependent Variable?’, In J. Tommola (ed) Topics in Interpreting Research , Turku: Centre for Translation and Interpreting, University of Turku, 121-133.
User Responses to Simultaneous Interpreting ANNA-RIITTA VUORIKOSKI University o f Tampere, Finland User responses to simultaneous interpreting (SI) is a relatively ne glected area in interpreting studies. The following article explores the nature o f user responses to simultaneous interpreting and asks whether there are unifying features in the diversity o f responses to SI. The article is based on the findings o f a survey carried out in five different seminars where the languages were English and Fin nish. The participants came from different parts o f Finland and represented many different occupations and educational back grounds. The article highlights the heterogeneity o f audiences and the necessity fo r more exact definition o f quality criteria in inter preting research. The importance o f informed interpreting , clear definition o f the overall communicative context and shared infor mation are further issues that are discussed.
Introduction Simultaneous interpreting (SI) refers to an immediate oral translation of spoken texts. The process itself has puzzled researchers, interpreters and listeners alike. The technical and communicative aspects are also frequently taken up by the conference participants in their discussions with the inter preters, and even more frequently by the interpreters among themselves. One of the questions that is often raised in these professional debates is conference participants’ use of SI. For example, Finnish interpreters won der why many conference participants do not listen to the interpreting, particularly if the languages are English and Finnish. Interpreters suspect that Finns may find it embarrassing to have to resort to SI. It is not unu sual, for example, that those members of the audience who are wearing the headsets tend to sit at the back of the room. Interpreters are aware, however, that language skills may not be the only factor that influences the use or non-use of SI. The members of the audience are experts in the field that is discussed in the seminar, and consequently their knowledge of the topic and of the specialized vocabulary may enable them to follow the original speaker without interpreting. Seminar participants may also come to the conclusion that they cannot trust the interpreters if they find that the interpreters do not demonstrate enough expertise in the topic that is being discussed in the meeting. Thus the reason for not making use of SI could be the inadequate quality of the interpreting. Looking at this far from complete list of opinions, attitudes and ap-
188
Unity in Diversity
proaches, is it at all possible to find any unity in the diversity of audience responses?
Aims, hypothesis and method With the above questions from the realities of working life in mind it seemed tempting to investigate to what extent mediated communication, such as is carried out in the simultaneous mode, is satisfactory to listen ers. For the purposes of the thesis that has been used as the source for this article (Vuorikoski 1995a) it was hypothesized that communication in the SI mode is satisfactory if the listeners use the service, and if they give positive feedback on the content and quality of the oral translation they have received. The study was carried out by consulting five seminar audi ences about the usefulness and quality of the service in meetings where SI was provided. Answers were elicited with the help of a questionnaire and supplementary interviews. Quantitative methods were used to organize and describe the response contained in the 173 questionnaires that were returned out of the 480 that were handed out. Such an approach is in compliance with the principles of modem translation theories (e.g. Reiß and Vermeer 1984, Hatim and Mason 1990) as they emphasize the func tion and purpose of translation, which means that the translator/interpreter needs information about the recipients and their needs and expectations. The survey method used in the study was selected to investigate the con ference interpreter’s target groups and the function and purpose of SI for these groups.
Discussion Interpreters * target groups
The discussion of SI tends to take it for granted that there is the SI user, or, as Kurz (1989, 1994) and others (Biihler 1986, Gile 1990) have hy pothesized, a more or less homogeneous group of users whose SI quality needs and expectations remain constant. These authors have hypothesized that SI users have certain expectations toward SI that are determined by the topic and type of meeting. Kurz (1994:4) suggested that different user groups may attribute different weight to different quality criteria. Kurz based her study on eight quality criteria out of the 15 that had first been used by Biihler for evaluating interpreters. The criteria selected by Kurz were: native accent, pleasant voice, fluency of delivery, logical cohesion of utterance, sense consistency with original message, completeness of interpretation, correct grammatical usage, and use of correct terminol ogy. A very careful reading of Kurz’s conclusions reveals that while there
Vuorikoski: User Responses to Simultaneous Interpreting
189
was agreement among the groups about the importance of some of the criteria, there was also variance among the individual groups about the importance of the other criteria (the groups represented participants in an international conference on general medicine, on quality control, and a Council of Europe meeting). Thus, in her concluding sentence Kurz (1994:6) returns to the singular in saying that “the target language re ceiver or listener must be seen as an essential element in the [interpreting] process”. It can be inferred from Kurz’s article that differences in group preferences are not of sufficient significance to warrant a change in the professional standards or practice. Consequently, the results and conclu sions of her studies did not lead to any modifications of existing SI theories either. The more recent AIIC study (Moser 1995) also investigated the rank ing of various SI quality criteria by conference delegates whose response was analyzed against the type of conference they participated in. (The survey on user expectations included elements such as ‘completeness of rendition’, ‘clarity of expression’, ‘correct terminology’; ‘focus on essen tials’ as opposed to ‘a complete rendition’, and ‘faithfulness to the meaning’ as opposed to a ‘literal reproduction of what is said’). Participants in dif ferent types of conferences did not express a clear preference for any given criterion; instead, the alternatives supplied by the interviewer found more or less equal support from the respondents. What had become evi dent in the study by Kurz was also the main conclusion in Moser’s study: the large majority stated that the most important quality criterion for SI is faithfulness to the meaning of the original. A professional interpreter might respond: Obviously, what else would the member of the audience expect of a conference interpreter? The findings of the empirical studies cited and those of the present study indicate that although the organizers have specified the target group of the meeting, it cannot be assumed that individual audiences are homo geneous; consequently the needs and expectations of the individuals within an audience vary considerably. In a seminar audience there will be addressees who want to understand the presentation in all its nuances. For this reason, these addressees may decide to listen to the source language addressor, but this may also be the reason to listen to SI, if the addressees do not have a sufficient competence of the foreign language in order to be able to follow the original presentation. Moreover, while one half of the audience may prefer to receive the essentials of the message, and for them logical cohesion of utterance and fluency may be the most important cri teria, there will be a considerable number of addressees who, often for professional reasons, want a complete rendering and a correct use of terms. The above findings suggest that the speech situation has to be analyzed carefully (cf. Pochhacker 1994), and a sufficient amount of information
190
Unity in Diversity
about the SI users has to be available, before one can conclude anything definite about the SI users’ needs and expectations, either as individuals or as groups. The results obtained for the present article (cf. Vuorikoski 1995b) in dicated that the SI users’ expectations may not be meeting or group specific. The results also highlighted the fact that the users’ SI needs did not re main constant throughout the meeting, but they varied, on one hand, depending on the SI receiver’s personal characteristics, and on the other hand, depending on certain features characterizing the address. In a semi nar-type of meeting where some or most of the addressors spoke English and the addressees were Finns, there seemed to be at least three types of categories of addressees: one category did not need SI because they were able to follow the presentations in English. The addressees in the second category listened to both the original and the interpreting in parallel, or, alternatively, they listened to the original, and if they did not understand something, or if they did not know a word, they listened to the interpret er’s version. The interpreter thus fulfilled the reassurance needs of these addressees, who used SI as a support. The third category consisted of addressees who were either dependent on SI because they did not know English well enough, or who had decided that SI worked well for them and saved them the effort of having to concentrate on understanding both the language and the message. The seminar participants did not necessar ily stay in the same category throughout a given meeting; instead, they may move from one category to another, depending on their mood or interests or the hour of the day or, most importantly, depending on the addressor’s idiosyncratic features and presentation style. Quality criteria
While the method used in the study referred to evinced a sufficient amount of information for outlining the interpreter’s typical target groups in one type of meeting, it did not provide precise evaluations of the quality of SI in the seminars concerned. Questions of SI quality had been the underly ing motivation for carrying out the study, but as the study progressed, it became increasingly evident that in order to serve their purpose, quality concepts have to be clearly defined - a fact that is well known in the industrial world. The quality criteria that had been used by other scholars and that were adapted for the present study could be understood in sub jective ways. Previous surveys did not provide definitions for the SI quality criteria that they had used, such as “fluency of delivery”, “sense consistency with the original message”, “logical cohesion of utterance”, “completeness of interpretation” (Kurz), or “faithfulness to the meaning of the original”, “literal reproduction of what is said”, “completeness of
Vuorikoski: User Responses to Simultaneous Interpreting
191
rendition”, “focus on essentials”, etc. (Moser). Such phrases and con cepts may seem self-explanatory to the experts discussing translation and interpreting, and respondents filling in questionnaires may understand them more or less the way intended by the surveyor. In effect, they may be sufficient for clarifying what SI users expect in general. For the purposes of SI studies, however, it is necessary to lay a proper theoretical basis for the quality criteria that are explored. The main problem underlying the fuzziness of the quality criteria that have been referred to above is the fact that they were not based on a theoretical foundation that would have endowed them with more sub stance (cf. Mack and Cattaruzza 1996:47). As an example of the problematic nature of the quality criteria used in the earlier empirical stud ies one may take ‘completeness of rendition’ which was contrasted with ‘focus on essentials’. They would deserve a study of their own, the analy sis being based on an appropriate theoretical approach and a sound method, such as a propositional analysis. In the AIIC questionnaire ‘completeness of rendition’ was used to refer to the same idea as ‘a literal reproduction of what is said’ without any reference either to the linguistic or rhetorical implications, or to communication theory. One of the conclusions by the present author is that each term and concept has to be defined carefully if the aim of a further study is to obtain more specific results than what has been achieved so far, and even tually develop something that can be considered relevant for interpreting theory. A study of the user groups and the communicative context should either start or end by defining what is meant by a user group and the communicative context (cf. Jansen 1996:12-13). If different user groups are discussed without elaborating what constitutes a user group, the con clusions will remain rather vague. A reader may infer by implication only that it is either the conference delegates’ education or the topic of the conference that constitutes a user group. The conclusion derived from the analysis of the five different audiences discussed here was that while it is important for the interpreter to have the end user in mind, there are other aspects in the communicative situation than the topic of the meeting or the education/occupation of the addressees that determine the use of and expectations toward SI. This became evident when analyzing the evaluation of the SI by the respondents in the five seminars that were surveyed. The first key finding was that in one seminar where the interpreting sounded excellent to the experienced colleague in the audience there was still some variance in the evaluation. Besides pointing at the ambiguity of the evaluation criteria the result proved the subjectivity of all quality evaluation. This finding should be taken into account when developing interpreter testing, for ex ample.
192
Unity in Diversity
Another key finding which was connected with the previous one was made while analyzing the evaluation of SI by the respondents in two semi nars which had the same organizer and the same topic and purpose, and where the target groups of the seminars had been defined using the same terms. The interpreters in the two seminars were also the same, yet the use of SI and the evaluations of it differed considerably between the two audiences. The respondents in the latter seminar were much more critical than those in the first one. In an attempt to explain the variance in the evaluations, the addressee characteristics that differentiated the two audi ences were analyzed. In the seminar with a more critical response the average age of the respondents was fairly low (+/- 40), and about one half of the participants were women as compared with the first seminar where the average age was considerably high (50+), and only ten per cent were women. There were no significant differences in the educational back ground of the addressees in these two seminars. The interpreters had received very little prior information about the seminars which took place on two consecutive days. Thus the differences in the evaluation could not be explained through a change in the quality of interpreting. The response seems to reflect the respondents’ individual expectations and also their response to the primary addressors’ presentations. This is all the more likely since the one clear difference between the two seminars was the list of speakers. Except for one lecturer, all the speakers changed from one day to the next. The results point to the need to analyze the addressors’ texts more closely and to compare them with the interpreters’ renderings in order to explore a few selected criteria in greater detail. This is a con clusion that other scholars in the field of SI studies have reached as well (e.g. Pochhacker 1995, Sunnari 1995). ‘Sense consistency with the origi nal’, or ‘faithfulness to the meaning of the original’ were the obvious criteria that listeners set for interpreting in previous studies. Furthermore, ‘accuracy’ was one of the criteria that caused differences in the evalua tion by the five audiences analysed in the present article. Quality criteria such as these need to be specified with reference to the context as well as both the source language and target language texts. Such a proposition could be further justified by the intriguing finding among the responses according to which the criticism of the SI in a given seminar often re ferred to features in the communication situation that were beyond the interpreters’ control. Many of these features were linked with the source language address and reflected cultural differences between the addres sors and the addressees. ♦ According to the response obtained through the survey that has been reported here, the most important criterion out of the six supplied in the questionnaire that was used in the present study proved to be that inter preting should be informed. In all the five seminars the majority of the
Vuorikoski: User Responses to Simultaneous Interpreting
193
respondents considered that the interpreting had been informed. How ever, a few spontaneous comments called for collaboration between the organisers and the interpreters to guarantee that the interpreters are in formed about the topic of the meeting and that they will acquire the terminology of the specialized field. The main actors in the communica tion situation that was analyzed in the present study are the organizers, the addressors, the addressees and the interpreters, who are both ad dressees and addressors. The organizer(s), the primary addressor(s), the addressees and the interpreters come from a diversity of backgrounds and share different amounts of knowledge about the topic and purpose of the meeting. In the meeting they are united through the texts. Furthermore, some members in the audience are united through SI technology and the interpreters by listening to their oral translations of the conference texts. In an ideal situation, all the actors are sufficiently informed about the topic and the purpose of the communicative situation and of each actor’s role there.
Conclusion The results of the present study indicate that collaboration is needed be tween all the parties involved in order to reach a communicative situation that is satisfying to all the parties involved. Communication could be en hanced if the speakers, and the organizers, too, understand the special demands of the situation. In an ideal situation each actor is aware of the others’ roles and takes them into consideration. If each actor is sufficiently informed about the situation, the topic, and the predominant expectations of the other actors, the resulting performance by the interpreters, i.e. the interpreters’ text, will be informed as well, and the overall quality will be closer to the users’ needs and expectations. Thus, while there is a great deal of diversity among the SI users’ needs and expectations, some unity will be introduced by the shared information and background knowledge of all the parties in the communicative situation.
References Blihler, H. (1986) ‘Linguistic (Semantic) and Extralinguistic (Pragmatic) Cri teria for the Evaluation of Conference Interpretation and Interpreters, Multilingua 5(4):231-235. Gile, D. (1990) ‘L’évaluation de la qualité de d’interprétation par les délégués: une étude de cas’, The Interpreters Newsletter 3:66-71. Hatim, B. and I. Mason (1990) Discourse and the Translator , London & New York: Longman.
194
Unity in Diversity
Jansen, P. (1996) The Role of the Interpreter in Dutch Courtroom Interac tion: the Impact of the Situation on Translational Norms’, in Jorma Tommola (ed) Topics in Interpreting Research , University of Turku: Centre for Translation and Interpreting, 11-36. Krawutschke, P. (ed) (1995) Connections, Proceedings of the 36th Annual Conference of the American Translators Association, Medford, N.J.: In formation Today, Inc. Kurz, I. (1989) ‘Conference Interpreting - User Expectations’, in Deanna Lindberg Hammond (ed) Proceedings o f the 30th Annual Conference o f the American Translators Association , Medford, NJ: Learned Informa tion, Inc., 143-148. Kurz, I. 1994. ‘What do different user groups expect from a conference inter preter?’, The Jerome Quarterly 9(2):3-6. Mack, G. and L. Cattaruzza (1996) ‘User Surveys in SI: a Means of Learning about Quality and/or Raising some Reasonable Doubts’, in Jorma Tom mola (ed) Topics in Interpreting Research , University of Turku: Centre for Translation and Interpreting, 37-49. Moser, P. (1995) Survey on Expectations o f Users o f Conference Interpreta tion. Final Report January 1995 , Commissioned by AIIC International Association of Conference Interpreters, Produced by SRZ Stadt + Regionalforschung GmbH, Vienna, Austria. Pochhacker, F. (1994) Simultandolmetschen als komplexes Handeln (Lan guage in Performance 10), Tübingen: Gunter Narr. Pöchhacker, F. 1995. Slips and shifts in simultaneous interpreting, in Jorma Tommola (ed) Topics in Interpreting Research , University of Turku: Centre for Translation and Interpreting, 73-90. Sunnari, M. (1995) ‘Processing Strategies in Simultaneous Interpreting: Ex perts vs. Novices’, in P. Krawutschke (ed) Connections, Proceedings of the 36th Annual Conference of the American Translators Association, Medford, N.J.: Information Today, Inc., 157-164. Tommola, Jorma (ed) (1996) Topics in Interpreting Research , University of Turku: Centre for Translation and Interpreting. Vuorikoski, A-R. (1995a) Audience Response to Simultaneous Interpreting. Unpublished licentiate thesis, University of Tampere, Department of Trans lation Studies. Vuorikoski, A-R. (1995b) ‘Simultaneous Interpreting as Experienced by the Audience, in P. Krawutschke (ed) Connections, Proceedings of the 36th Annual Conference of the American Translators Association, Medford, N.J.: Information Today, Inc., 165-174.
List of Contributors Paul Bennett
Centre for Computational Linguistics, UMIST PO Box 88, Manchester M60 1QD, United Kingdom, paul@ cclum ist.ac.uk Ian Brown
Queen Margaret College Corstorphine Campus, Clerwood Terrace Edinburgh EH 12 8TS, Scotland Louise von Flotow
School of Translation and Interpretation, University of Ottawa PO Box 450, Ottawa, Ontario KIN 6N5, Canada. vanflotow@ aixl.uottawa.ca Marketta Heleva
Centre for Translation and Interpreting, University of Turku Tykistokatu 4, FIN-20520 Turku, Finland Matthias Heyn
TRADOS Benelux S.A. 303 avenue de Tervueren, 1150 Brussels, Belgium, [email protected] Susan Ingram
Comparative Studies, University of Alberta Edmonton, AB T6G 2E3, Canada. [email protected] Irena Kovacic
Faculty of Arts, University of Ljubljana 61001 Ljubljana, Askerceva 2, Slovenia Sara Laviosa
Department of Language Engineering, UMIST P O Box 88, Manchester M60 1QD, United Kingdom, [email protected] Magnus Merkel
NLPLAB, Department of Computer and Information Science Linköping University, Sweden, [email protected] Sharon O'Brien
ALPNET Ireland Ltd. 17-19 Sir John Rogerson’s Quay, Dublin 2, Ireland, [email protected] Eithne O'Connell
SALIS, Dublin City University Dublin 9, Ireland. [email protected]
196
Unity in Diversity
Carol Peters
Istituto di Elaborazione della Informazione Consiglio Nazionale delle Ricerche Via S. Maria, 46, 56126 Pisa, Italy, carol@ iei.pi.cnr.it Eugenio Picchi
Instituto di Linguistica Computazionale Consiglio Nazionale delle Ricerche Via della Faggiola 32, 56126 Pisa, Italy, picchi@ ilc.pi.cnr.it Franz Pöchhacker
Department of Translation and Interpreting, University of Vienna Gymnasiumstr. 50, A-1190 Vienna, Austria [email protected] Christina Schäffner
Department of Languages and European Studies, Aston University Aston Triangle, Birmingham B4 7ET, United Kingdom [email protected] Reinhard Schäler
Localisation Resources Centre, Roebuck Castle, University College Dublin Belfield, Dublin 4, Ireland. [email protected] Ceri Sherlock
8 Kyveilog Street, Cardiff CF1 9J1, Wales, United Kingdom Anikó Sohdr
CETRA Research Centre for Translation, Communication and Cultures Katholieke Universiteit Leuven Blijde-Inkomststraat 21, 3000 Leuven, Belgium [email protected]. kuleuven. ac. be Paul St-Pierre
Département de linguistique et de traduction, Université de Montréal C.P. 6128, Succursale Centre-ville, Montréal H3C 3J7, Canada saintpip @ ere. umontreal. ca Jorma Tommola
Centre for Translation and Interpreting, University of Turku Tykistokatu 4, FIN-20520 Turku, Finland, [email protected] Anna-Riitta Vuorikoski
Onkiniemenkatu 1C 22, FIN-33230 Tampere, Finland Judy Wakabayashi
Department of Asian Languages and Studies, University of Queensland Brisbane 4072, Australia. [email protected]