Symposium on Lexicography V: Proceedings of the Fifth International Symposium on Lexicography May 3–5, 1990 at the University of Copenhagen 9783111341095, 9783484309432

Citation preview

mmmQE^pmcA S;";

LEXICOGRAPHICA Series Maior Supplementary Volumes to the International Annual for Lexicography Suppléments à la Revue Internationale de Lexicographie Supplementbände zum Internationalen Jahrbuch für Lexikographie

Edited by Sture Allén, Pierre Corbin, Reinhard R. K. Hartmann, Franz Josef Hausmann, Hans-Peder Kromann, Oskar Reichmann, Ladislav Zgusta

43

Published in cooperation with the Dictionary Society of North America (DSNA] and the European Association for Lexicography (EURALEX)

Symposium on Lexicography V Proceedings of the Fifth International Symposium on Lexicography May 3-5, 1990 at the University of Copenhagen edited by Karl Hyldgaard-Jensen and Arne Zettersten

Max Niemeyer Verlag Tübingen 1992

Die Deutsche Bibliothek - CIP-Einheitsaufnahme Symposium on Lexicography : Symposium on Lexicography V : proceedings of the Fifth International Symposium on Lexicography May 3 - 5 , 1990 at the University of Copenhagen / ed. by Karl Hyldgaard-Jensen and Arne Zettersten. - Tiibingen : Niemeyer, 1992 (Lexicographica : Series maior ; 43) NE: Hyldgaard-Jensen, Karl [Hrsg.]; K0benhavns Universitet; Lexicographica / Series maior ISBN 3-484-30943-1

ISSN 0175-9264

© Max Niemeyer Verlag GmbH & Co. KG, Tübingen 1992 Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Printed in Germany. Druck: Weihert-Druck GmbH, Darmstadt Einband: Hugo Nädele, Nehren

TABLE O F CONTENTS

Introduction English French German

1 7 13

Joachim M u g d a n On the Typology of Bilingual Dictionaries

17

Hans Kristian Mikkelsen W h a t did Sierba actually m e a n by "active" and "passive" Dictionaries?

25

Ulrich Heid Syntactic Information in (Machine) Translation Dictionaries - Towards a Modular Architecture for Bilingual Dictionaries

41

Klaus Schubert Lexicography, or Corpus-based Knowledge Acquisition?

81

Roda P. Roberts Methods of Bilingual Dictionary-Making: The Canadian Experience

91

Ilse Karl Ist die gleichartige Gestaltung von allgemeinen Bedeutungswörterbüchern verschiedener Sprachtypen erstrebenswert?

117

Albertas Steponavicius Basic Principles of an English-Lithuanian Dictionary of Idioms

131

Andreas F. Kelletat Wie k o m m e n die Wörter ins Wörterbuch? Zur Makro-struktur zweisprachiger Wörterbücher am Beispiel des Sprachenpaars Finnisch/Deutsch

143

Gregor Meder Zum Stand der deutsch-(serbo)kroatischen/ (serbo)kroatisch-deutschen Lexikographie: Die wichtigsten Wörterbücher

159

Norman Shine Dictionaries for Specialised Languages. The Situation in Denmark Today

175

VI Baidur Panzer Sellius' Vocabularium Russo-Germanicum und die frühe deutsch-russische Lexikographie

189

Sven-Göran Malmgren German-Swedish Lexicography in 18th Century Pomerania

201

Jorgen Erik Nielsen Svend Rosing, his "Engelsk-Dansk Ordbog" (1853), and Gyldendal

217

Inge Kabeil / Hanne Lauridsen Charles Bertram (1723-1765): Language Teacher, Lexicographer and Impostor

227

Dieter Herberg Antonymische Beziehungen im Wortschatz und im Wörterbuch

245

Edward Gates Should a Dictionary Include only the "good" Words? ...

265

Gunnar Persson and its Friends: Some Preliminary Observations False

281

Bernhard Diensberg Lexical Change in the "Ancrene Riwle", with Special Consideration of the Romance and Scandinavian Loanwords

295

Jens Eric Mogensen Zur lexikographischen Beschreibung einiger inhaltsarmer Verben im Deutschen

315

Anneliese Möller Zum Anteil von Wörterbüchern und Wörterverzeichnissen an der Ausprägung und Kodifizierung einer einheitlichen graphischen Norm des Deutschen

331

Ulrich Busse Doppelformen von Anglizismen im Rechtschreib-DUDEN ...

341

HansOn Peters the Historical Development of English Boosters ....

373

Viggo Hjifrnager Pedersen Center for Translation Studies. A Presentation

393

The Symposium on Lexicography 1990. List of Participants

397

APPENDIX

Broder Carstensen Die Lexikographie in der Informationsgesellschaft ....

401

VII Broder Carstensen Die Bedeutung der Lexikographie in der Informationsgesellschaft

402

Walter V o i g t Einige Bemerkungen zur zweisprachigen Verlagslexikographie

405

Franz Josef Hausmann Die zweisprachigen Wörterbücher

408

Veronika Schnorr Lexikographie zwischen Theorie und Praxis

411

Herbert Ernst Wiegand Einsprachige Gebrauchswörterbücher - Verlagslexikographie - akademische Wörterbuchforschung. Plädoyer für einen Diplomstudiengang zur Lexikographie

414

Günther Drosdowski Traum und Wirklichkeit

418

Karl Hyldgaard-Jensen Zusammenfassende Stellungnahme aus der Sicht eines ausländischen Germanisten

422

INTRODUCTION

The F i f t h I n t e r n a t i o n a l v e r s i t y of C o p e n h a g e n ,

S y m p o s i u m o n L e x i c o g r a p h y a t the U n i took p l a c e on M a y 3 - 5 , 1990. T h e

t i c i p a n t s came f r o m A u s t r i a , B u l g a r i a , C a n a d a , Finland, France,

par-

Denmark,

the F e d e r a l R e p u b l i c of G e r m a n y ,

the G e r m a n

Democratic Republic, Great Britain, Iceland, India,

Lithuania,

the N e t h e r l a n d s , N o r w a y , a n d S w e d e n . 30 p a p e r s w e r e r e a d in sections chaired by Broder Carstensen, Arne Zettersten, - P e d e r K r o m a n n , Karl H y l d g a a r d - J e n s e n , V i g g o

Hans

Hjarnager

P e d e r s e n , D i e t e r H e r b e r g , U l r i c h Heid, E b b a H j o r t h , a n d H e n n i n g Bergenholtz. A t the m a i n c o n f e r e n c e d i n n e r , P r o f e s s o r B r o d e r P a d e r b o r n , w a s e l e c t e d H o n o r a r y m e m b e r of the

Carstensen,

Copenhagen

Symposium. T h e m a i n p a r t of the p r o g r a m m e w a s this time d e v o t e d to p r o b lems c o n n e c t e d w i t h b i l i n g u a l (University of Münster) lecture,

lexicography. Joachim

Mugdan

o p e n e d the s y m p o s i u m w i t h a p l e n a r y

"On the T y p o l o g y of B i l i n g u a l D i c t i o n a r i e s " . He

c i z e d the d i f f e r e n t i a t i o n of d i c t i o n a r i e s h i t h e r t o p l e a d i n g for an e x t e n s i o n of the list of d i c t i o n a r y to c o v e r all four c l a s s i c a l

functions

skills and multifunctional

t i o n a r i e s w h i c h a c t u a l l y c a t e r for the w i d e s t p o s s i b l e of u s e r n e e d s . A n E n g l i s h s u m m a r y is p r i n t e d in this the c o m p l e t e text "Zur T y p o l o g i e z w e i s p r a c h i g e r in M e d e r , G r e g o r / D ö r n e r , A n d r e a s terbücher

criti-

adopted dicvariety

volume,

Wörterbücher",

(eds.): W o r t e , - W ö r t e r - W ö r -

( L e x i c o g r a p h i c a S e r i e s Maior),

Hans Christian Mikkelsen

(Aarhus B u s i n e s s School)

discusses

m e a n i n g of the terms " a c t i v e " a n d " p a s s i v e " d i c t i o n a r i e s

as-

the

2

Vy cribed to the Russian linguist Scerba. He concludes that *v Scerba's concept of explanatory vs. translation dictionaries differs from the present distinction between VW active and passive dictionaries, which terms are not used by Scerba in connection with dictionaries. Ulrich Heid (University of Stuttgart) deals with much more than the title of his contribution "Syntactic Information in (Machine) Translation Dictionaries - towards a Modular Architecture for Bilingual Dictionaries" indicates. He surveys general problems of bilingual lexicography and argues for a modulary system with typed feature structures developed by the Polygloss project, Bonn. As an example he describes the hierarchical lexical organisation of the syntactic properties of lexemes. Klaus Schubert (BSO/Research, Utrecht) presents an approach to knowledge aquisition for machine translation purposes in which he attempts to automate part of the lexicographer's work. He outlines an application of this technique for an advanced form of computer aided lexicography. Roda Roberts (University of Ottawa) outlines the methodological principles adopted in the Canadian project for a bidirectional bilingual English/French dictionary reflecting the Canadian usage. U s e Karl (Akademie der Wissenschaften, Berlin) tackles the delicate problem of using the same principle of description in dictionaries for languages belonging to different types of language, such as German and Chinese. Albertas Stepanovicius (University of Vilnius) describes the history of multilingualism in Lithuania and the development of English-Lithuanian dictionary making, as well as the basic principles for a planned English-Lithuanian dictionary of idioms.

3

Andreas F. Kelletat (University of Vasa) discusses the distinction between linguistic and encyclopaedic dictionaries. In his opinion this distinction cannot be maintained strictly in a bilingual dictionary but it is a question how much encyclopaedic material should be included in the bilingual dictionary. Gregor Meder (University of Essen) reviews existing German(Serbo-) Croatian dictionaries pointing out special deficiencies in German/(Serbo-) Croatian lexicography to be remedied in future dictionaries of the kind. Norman Shine (University of Copenhagen) surveys the situation in Denmark as regards dictionaries for special purposes. Some papers dealt with older bilingual lexicography. In connection with his publication of "Sellius1 Vocabularium Russo-Germanicum", Baldur Panzer (University of Heidelberg) surveys early German-Russian lexicography, whilst Sven-GOran Malmgren (University of Gothenburg) traces the influence of early German, especially Low-German lexicography, on the first SwedishGerman, dictonaries in 18th-century Pomerania. Older Danish lexicographers are treated by Jorgen E. Nielsen (University of Copenhagen), who describes the genesis of the famous Gyldendal dictionaries, Danish-English and English -Danish, dealing at length with the works of Svend Rosing, and by Inge Kabell/Hanne Lauridsen (University of Copenhagen) who investigate the background of the Danish-English lexicographer Charles Bertram. The second part of the programme comprised papers dealing with problems of general lexicographic interest. Dieter Herberg (Akademie der Wissenschaften, Berlin) draws attention to the importance of antonymic relations in language and in lexicography and offers a suggestion as to how to im-

4 prove the systematic description of these relations in dictionaries . In "Should a Dictionary include only the "Good" Words?", Edward Gates (University of Regensburg) discusses the conflicting views as to the inclusion of (potentially) morally objectionable words in general monolingual dictionaries. As an alternative to the traditional lexicographic method of paraphrasing words as false in terms of near-synonyms, Gunnar Persson (University of Umeâ) recommends a classification based on the gestalt analysis adopted by Lakoff and Johnson. Berhard Diensberg (University of Tromsö) examines how far lexical change in Middle English has been taken into account and recorded in MED and OED. Jens E. Mogensen (University of Copenhagen) criticises in his contribution "Zur lexikographischen Beschreibung inhaltsarmer Verben im deutschen" the presentation of these verbs in valence dictionaries of verbs, finding that they should be excluded and instead treated in valence dictionaries of nouns. Anneliese Möller (University of Rostock) examines the role dictionaries have played for the establishment of a uniform German orthography. Ulrich Busse (University of Paderborn) describes in "Doppelformen von Anglizismen im Rechtschreib-Duden" the changing ratio

of these forms from the end of the 19th century

to the present day. Hans Peters (Freie Universität, Berlin) scrutinizes the historical development of English boosters, a special category of degree adverbs. He states that alphabetically organized dictionaries are of little help and looks forward to the publication of the Historical Thesaurus of the English language.

5 Viggo Hjarnager Pedersen

(University of Copenhagen) presents

the activities of the Center for Translation Studies and Lexicography at the University of Copenhagen. In the appendix "Die Bedeutung der Lexicographie in der Informationsgesellschaft", Broder Carstensen publishes the papers read at the 1990 conference of the German association AnGeRo (Anglisten-, Germanisten- und RomanistenverbSnde) by Broder Carstensen, Walter Voigt, Franz Josef Hausmann, Veronika Schnorr, Herbert Ernst Wiegand, Giinter Drosdowski und Karl Hyldgaard-Jensen. Acknowledgements: The editors wish to thank the authors of the contributions for placing their manuscripts at our disposal and all participants, old friends and newcomers, for joining the symposium. We are indebted for financial support to the Danish Research Council for the Humanities, Einar Hansen's Forskningsfond, the Center for Translation Studies and Lexicography, and the Faculty of the Humanities, Copenhagen University, and we cordially thank the Gyldendal Publishing House and the Ambassador of the Federal Republic of Germany in Denmark for grants to the social frame of the symposium.

Copenhagen, May 1990

The editors

INTRODUCTION

Le Cinquième Symposium International de Lexicographie organisé par l'Université de Copenhague s'est tenu du 3 au 5 mai 1990. Les participants venaient des pays suivants: Autriche, Bulgarie, Canada, Danemark, Finlande, France, Grande-Bretagne, Inde, Islande, Lituanie, Norvège, Pays-Bas, République Démocratique Allemande, République Fédérale d'Allemagne, Suède. Trente communications ont été faites dans des sections présidées par Broder Carstensen, Arne Zettersten, Hans-Peder Kromann, Karl HyldgaardJensen, Viggo Hjornager Pedersen, Dieter Herberg, Ulrich Heid, Ebba Hjorth et Henning Bergenholtz. La majeure partie du programme de ce symposium a été cette fois consacrée aux problèmes concernant la lexicographie bilingue. Le symposium s'est ouvert sur une séance plénière, où Joachim Mugdan (Université de Münster) a pris la parole. Dans sa communication, "On the Typology of Bilingual Dictionaries", le conférencier a critiqué la différenciation qu'on a établie jusqu'à présent entre les dictionnaires, et il a préconisé qu'on élargisse la liste des fonctions dictionnairiques de sorte qu'elle englobe les quatre vertus traditionnelles et des dictionnaires multifonctionnels qui répondent réellement aux besoins les plus variés des usagers. Un résumé en anglais est publié dans le présent volume. Le texte intégral, "Zur Typologie zweisprachiger Wörterbücher", figure dans Meder, Gregor/Dörner, Andreas (éd).: Worte - Wörter Wörterbücher (Lexicographica Sériés Maior). Hans Kristian Mikkelsen (Ecole Supérieure de Commerce de Aarhus) approfondit le sens des termes "dictionnaires actifs" et "dictionnaires passifs", dont on attribue la paternité au linguiste

8 V V

russe Scerba. Il conclut en disant que le concept, établi par ¡acerba, des dictionnaires "explicatifs" par opposition aux dictionnaires "traductifs" diffère de la distinction qu'on fait aujourd'hui entre dictionnaires actifs et dictionnaires passifs, termes que Scerba n'utilise pas en relation avec des dictionnaires . Ulrich Heid (Université de Stuttgart) va plus loin que ne l'indique le titre de sa communication, "Syntactic Information in (Machine) Translation Dictionaries - towards a Modular Architecture for Bilingual Dictionaries". Il se penche sur des problèmes d'ordre général en lexicographie bilingue et argumente en faveur d'un système modulaire à partir de structures caractéristiques développées par le projet Polygloss, Bonn. Comme exemple, il décrit l'organisation lexicale hiérarchique des propriétés syntaxiques des lexèmes. Klaus Schubert (BSO/Research, Utrecht) présente une approche d'acquisition de savoir pour les besoins de la traduction par machine, dans laquelle il s'efforce d'automatiser une partie du travail du lexicographe. Il esquisse une application de cette technique à une forme avancée de lexicographie assistée par ordinateur. Roda Roberts (Université d'Ottawa) définit les principes méthodologiques mis en oeuvre pour le projet canadien d'un dictionnaire bilingue bidirectionnel anglais-français, reflétant l'usage de la langue au Canada. lise Karl (Akademie der Wissenschaften, Berlin) aborde le délicat problème qui consiste en l'emploi du même principe descriptif dans des dictionnaires de langues appartenant à des types linguistiques différents comme l'allemand et le chinois. Albertas Stepanovicius (Université de Vilnius) relate d'abord l'histoire du plurilinguisme en Lituanie et le développement de la confection d'un dictionnaire anglais-lituanien, puis il

9

expose les principes fondamentaux d'un dictionnaire d'idiomes anglais-lituanien mis en projet. Andréas F. Kelletat (Université de Vasa, Finlande) traite de la distinction entre dictionnaires de langue et encyclopédies. Selon lui, cette distinction ne peut être strictement observée dans un dictionnaire bilingue. Mais il s'agit de savoir quelle quantité de matériaux encyclopédiques doit être incluse dans un dictionnaire bilingue. Gregor Meder (Université d'Essen) passe en revue les dictionnaires allemand-(serbo-)croate existants, en insistant sur des lacunes et des défauts spécifiques dans la lexicographie allemand- (serbo-) croate auxquels il faudra porter remède dans les futurs dictionnaires de ce genre. Norman Shine (Université de Copenhague) présente un tableau des dictionnaires de langues de spécialités au Danemark. Quelques communications ont porté sur la lexicographie bilingue dans le passé. En rapport avec sa publication du "Sellius' Vocabularium Russo-Germanicum", Baldur Panzer (Université de Heidelberg) se penche sur la toute première lexicographie allemand-russe, tandis que Sven-Gôran Malmgren (Université de Gôteborg) relève l'influence du vieux allemand, et particulièrement de la lexicographie basse allemande, dans les premiers dictionnaires suédois-allemand publiés en Poméranie au XVIII* siècle. D'anciens lexicographes danois sont passés en revue par Jorgen E. Nielsen (Université de Copenhague), qui relate la genèse des fameux dictionnaires danois-anglais et anglais-danois de chez Gyldendal pour terminer par les ouvrages de Svend Rosing, et par Inge Kabell et Hanne Lauridsen (Université de Copenhague), qui étudient le "background" du lexicographe dano-anglais Charles Bertram. La seconde partie du programme s'est composée de communications portant sur des problèmes d'intérêt général en lexicographie.

10

Dieter Herberg (Akademie der Wissenschaften, Berlin) attire l'attention sur l'importance des relations antonymiques dans la langue et la lexicographie, et il suggère la façon dont on peut étendre la description systématique de ces relations dans les dictionnaires. Dans "Should a Dictionary include only the "Good" Words?", Edward Gates (Université de Regensburg) expose les points de vue qui sont source de conflit lorsqu'il s'agit d'inclure des termes pouvant être moralement condamnables dans des dictionnaires généraux monolingues. Comme solution de rechange au traitement lexicographique traditionnel appliqué aux mots paraphrastiques du genre faux-amis en matière de quasi-synonymes, Gunnar Persson (Université de Umeà) préconise une classification reposant sur l'analyse gestaltienne adoptée par Lakoff et Johnson. Bernhard Diensberg (Université de Tromso) examine à quel degré l'évolution lexicale en moyen anglais a été prise en compte et enregistrée dans MED et OED. Dans sa communication, "Zur lexikographischen Beschreibung inhaltsarmer Verben im Deutschen", Jens E. Mogensen (Université de Copenhague) critique la présentation de ces verbes dans les dictionnaires valenciels de verbes, et il estime qu'ils devraient en être écartés pour être traités dans les dictionnaires valenciels de noms. Anneliese Möller (Université de Rostock) étudie le rôle qu'ont joué les dictionnaires dans l'établissement d'une orthographe unifiée en allemand. Ulrich Busse (Université de Paderborn), dans sa communication, "Doppelformen von Anglizismen im Rechtschreib-Duden", décrit l'évolution de ces formes depuis la fin du XIX* siècle jusqu'à nos jours.

11

Hans Peters (Freie Universität, Berlin) étudie de près le développement historique des "boosters" en anglais, une catégorie spéciale d'adverbes d'intensité. Il constate que les dictionnaires ordonnés alphabétiquement sont d'une aide minime et il attend avec impatience la publication du Historical Thesaurus of the English language. Viggo Hj0rnager Pedersen (Université de Copenhague) trace un tableau des activités du Centre de Théorie de la Traduction et de Lexicographie de l'Université de Copenhague. Dans l'appendice "Die Bedeutung der Lexikographie in der Informationsgesellschaft", Broder Carstensen (Université de Paderborn) rapporte les communications faites en 1990 à la conférence de l'association germanique AnGeRo (Anglisten-, Germanisten- und Romanistenverbände) par Broder Carstensen, Walter Voigt, Franz Josef Hausmann, Veronika Schnorr, Herbert Ernst Wiegand, Günther Drosdowski et Karl Hyldgaard-Jensen. Remerciements : Les éditeurs tiennent à remercier les auteurs de communications qui ont bien voulu leur confier leur manuscrit aux fins de publication ainsi que tous les membres du groupe lexicographique qui ont participé au symposium. Nous sommes redevables, pour leur soutien financier, au C.N.R.S. danois (Statens Humanistiske Forskningsràd), à la Fondation pour l'encouragement de la recherche Einar Hansen, au Centre de Théorie de la Traduction et de Lexicographie de l'Université de Copenhague, et à la Faculté des Lettres de cette même université. Nous remercions cordialement la maison d'édition Gyldendal et Son Excellence, l'Ambassadeur de la République Fédérale d'Allemagne, d'avoir, par leur générosité, assuré le cadre social à ce symposium.

Copenhague, mai 1990

Les éditeurs

VORWORT

Das 5. internationale Symposium über Lexikographie der Universität Kopenhagen fand in der Zeit vom 3. bis 5. Mai 1990 mit Teilnehmern aus Bulgarien, Deutschland, Dänemark, Finnland, Frankreich, Grossbritannien, Island, Kanada, Litauen, den Niederlanden, Norwegen, Österreich und Schweden statt. 30 Referate wurden in zwei Sektionen gehalten. Im Zentrum des Interesses stand diesmal die zweisprachige Lexikographie. Joachim Mugdan eröffnete das Symposium mit dem Plenarvortrag "On the Typology of Bilingual Dictionaries", in dem er die bisher benutzte Differenzierung der Wörterbücher kritisierte und für eine Erweiterung der Zahl der WörterbuchFunktionen plädierte, die sämtliche vier klassische Fertigkeiten einschliessen würde, und für multifunktionelle Wörterbücher, die tatsächlich die grösstmögliche Berücksichtigung der Benutzerbedürfnisse gewährleisten, eintrat. Hans Kristian Mikkelsen untersucht eingehend die Bedeutung der beiden Termini "aktive" und "passive" Wörterbücher, die dem rus* * sischen Linguisten Scerba zugeschrieben werden. Er gelangt zu V v

der Schlussfolgerung, dass die Differenzierung Scerbas zwischen "erklärenden" und "übersetzenden" Wörterbüchern nicht mit der heute gebräuchlichen Distinktion zwischen "aktiven" und "passiven" Wörterbüchern zusammenfällt. Der Beitrag Ulrich Heids "Syntactic Information in (Machine) Translation Dictionaries - towards a Modular Architecture for Bilingual Dictionaries" umfasst weit mehr als das was der Titel andeutet. In Wirklichkeit durchmustert Heid hier die ganze Problematik der zweisprachigen Lexikographie, wobei er für die Einführung eines Modulsystems mit "typed feature structures",

14

das von dem Polygloss Projekt in Bonn entwickelt wurde, argumentiert. Zur Illustration des Systems beschreibt er den hierar chischen lexikalischen Aufbau der syntaktischen Eigenschaften von Lexemen. In "Lexicography or Corpus-based Knowledge Acquisition" schildert Klaus Schubert eine im BSO/Research, Utrecht entwickelte Methode zum Kenntniserwerb mittels des Com puters, die darauf hinausläuft, die Arbeit des Lexikographen op timal zu automatisieren. Roda Roberts gibt eine Übersicht über die methodischen Prinzipi en, die dem kanadischen Projekt "The Bilingual Canadian Dictionary", das den kanadischen Gebrauch des Englischen und des Fran zösischen kontrastiv verzeichnet, zugrundegelegt wurden. Ilse Karl fragt, ob man bei grosser typologischer Divergenz der beiden zu vergleichenden Sprachen - in casu Deutsch und Chinesisch - im bilingualen Wörterbuch die gleichen Beschreibungsprinzipien verwenden kann. Albertas Stepanovicius skizziert zum einen die Geschichte der litauischen Lexikographie, zum anderen die Hauptprinzipien für ein geplantes englisch/litauisches idiomatisches Wörterbuch, während Andreas F. Kelletat das Problem der Berücksichtigung enzyklopädischen Stoffes in dem linguistischen bilingualen Wörterbuch diskutiert Gregor Meder durchmustert die vorhandenen deutsch-(serbo)kroati sehen Wörterbücher, wobei er auf gewisse Mängel aufmerksam macht, denen in künftigen Wörterbuchprojekten abgeholfen werden sollte. Über den heutigen Stand der dänisch-englischen Fachwörterbücher in Dänemark berichtet Norman Shine. Eine Reihe von Referaten behandelten Teile der älteren zweispra chigen Lexikographie: In Verbindung mit seiner Edition von "Sei lius' Vocabularium Russo-Germanicum" gibt Baidur Panzer eine Übersicht über die frühe deutsch-russische Lexikographie, während Sven-Göran Malmgren dem Einfluss der frühen deutschen, besonders niederdeutschen Lexikographie auf die ersten deutschschwedischen Wörterbücher im Pommern des 18. Jahrhunderts nachgeht.

15

Ältere dänische Lexikographen behandeln Jörgen E. Nielsen, der die Entstehung der berühmten Gyldendal-Wörterbücher für Dänisch/Englisch beschreibt, und Inge Kabell/Hanne Lauridsen, die das wechselnde Schicksal des dänisch-englischen Lexikographen Charles Bertram verfolgen. Den zweiten Teil des Programms machten Referate, die Probleme der allgemeinen unilingualen Lexikographie aufgriffen, aus. Dieter Herberg macht auf die Rolle aufmerksam, die antonymische Beziehungen in der Sprache und in der Lexikographie spielen, und legt einen Vorschlag zur systematischen Beschreibung dieser Beziehungen in Wörterbüchern vor. In "Should a Dictionary include only the "Good" Words?" diskutiert Edward Gates die kontroversielle Frage, ob unanständige Wörter ins Wörterbuch aufgenommen werden sollen. Als eine Alternative zur traditionellen lexikographischen Methode, Adjektive wie eng. false durch sinnverwandte Wörter zu paraphrasieren, empfiehlt Gunnar Persson eine auf der von Lakoff & Johnson angewandten Gestaltanalyse basierende Klassifizierung. Bernhard Diensberg untersucht, inwieweit lexikalische Entwicklungen im Mittelenglischen von dem MED und OED berücksichtigt und registriert wurden. Jens E. Mogensen findet, dass die lexikalische Beschreibung inhaltsarmer Verben im Deutschen nicht in das Verbvalenzwörterbuch gehört, sondern in das Wörterbuch der Valenz der Substantive. Anneliese Möller weist nach, welche Rolle die deutschen Wörterbücher für die Heranbildung einer einheitlichen Orthographie des Deutschen gespielt haben, während Ulrich Busse in "Doppelformen von Anglizismen im Rechtschreib-Duden" die wechselnde Dominanz dieser Formen vom Anfang des 19. Jhs. bis heute beschreibt. Hans Peters analysiert die historische Entwicklung der sog. boosters im Englischen, und abschliessend stellt Viggo Hjörnager Pedersen das Center for Translation Studies and Lexicography der Universität Kopenhagen vor.

16

Im Appendix "Die Bedeutung der Lexikographie in der Informationsgesellschaft" publiziert Broder Carstensen die Referate, die 1990 auf der gleichnamigen Fachkonferenz der AnGeRo (Anglisten-, Germanisten- und Romanistenverbände Deutschlands) in Bonn gehalten wurden. Unter den Referenten waren Vertreter der Universitäten und der Verlage, die u.a. zu Themen wie Theorie und Praxis der Lexikographie, Verlagslexikographie versus Universitätslexikographie, Ausbildung der Lexikographen Stellung nahmen. Die Herausgeber danken sehr herzlich den Autoren für ihre Bereitschaft, die Manuskripte ihrer Beiträge für den Druck zur Verfügung zu stellen, sowie den Mitgliedern unserer Forschergruppe für ihre aktive Teilnahme an dem Symposium. Wir sind auch dem Dänischen humanistischen Forschungsrat, der Humanistischen Fakultät der Universität Kopenhagen und Einar Hansens Forskningsfond für die finanzielle Unterstützung der Veranstaltung sowie dem Gyldendal Verlag und dem Botschafter der Bundesrepublik Deutschland für Beiträge zu den Rahmenveranstaltungen sehr zu Dank verpflichtet.

Kopenhagen, im Mai 1990

Die Herausgeber

Joachim Mugdan ON THE TYPOLOGY OF BILINGUAL DICTIONARIES*

1.

"Active" vs. "Passive": A Reappraisal

In bilingual lexicography, it has become commonplace to contrast "active"

and

according

to

"passive" the

dictionaries

following

scheme

or (cf.

dictionary Lotzsch

&

functions al.

1983-

84:VII, Hausmann 1977:56-58, Kromann/Riiber/Rosbach 1984:185): dictionary | source lg.

target lg.

purpose

active passive

foreign native

production, encoding, Hinübersetzung reception, decoding, Herübersetzung

1 native | foreign

Table 1 This distinction is often attributed to Lev Vladimirovic Sfierba (1880-1944), who maintained that "for each pair of languages, four dictionaries are needed definitely two explanatory foreign-language dictionaries with explanations in the mother tongue of the user of the respective dictionary and, depending on actual needs, two dictionaries of a special kind [...] for translating from the mother tongue into the foreign language." (§£erba 1974 [1940]: 303, transl. J.M.)

* This is a revised summary of my contribution to the symposium. I have discussed the topic in more detail in "Zur Typologie zweisprachiger Wörterbücher", to appear in: Meder, Gregor / Dörner, Andreas (eds.), Worte - Wörter - Wörterbücher: Lexikographische Beiträge zum Essener Linguistischen Kolloquium 19821987, Tübingen: Niemeyer (Lexicographica Series Maior).

18 Sierba emphasized that a Russian-French dictionary for speakers of Russian must be quite different from one for French-speaking users,

but

what

he

had

in mind

dichotomy as we know it today to dictionaries kind for

can

be

from

identified

dictionaries

should

translation

dictionaries

-

lists

of

the

active/passive

"dictionaries of

the mother

explanatory opinion,

not

(and he did not apply these terms

either). While his

translating

language"

was

tongue into

with

the

not

be

to which

translation

active

type,

mistaken

§6erba

equivalents

a special

the foreign for

objected.

give

an

the

passive In

his

inadequate

impression of the true meaning of a word and therefore do not help the user to fully understand the foreign text. Ideally, he felt, the

student

of

a foreign

language

monolingual explanatory dictionary. the term explanatory dictionary work

which

provides

should

work

with

a

(In the Russian tradition,

[tolkovyj slovar'] refers to a

explanations

of

meanings;

the

word

definition is inappropriate in this context, cf. Wiegand 1985.) Sderba realized, however, that this might be too difficult for users with little competence in the foreign language. The idea of designing special monolingual dictionaries for learners did not occur to him; instead, he proposed a bilingual dictionary which explains the meanings of the foreign words in the user's mother

tongue

without

necessarily

offering

translation

equivalents. Essentially, these explanations could correspond to those

in

Deutsches the

a monolingual

dictionary.

(In

Bielfeldt's

Russisch-

Worterbuch of 1958, the best-known representative of

explanatory

type,

one

finds

many

translations

of

the

explanations in a monolingual Russian dictionary, Oiegov 1952.) According to Sfierba, "the fundamental rule of foreign language teaching methodology is never to translate from the mother tongue but to try and think in the foreign language as far as one's knowledge of it permits." (Sierba 1974 [1939]:307) This is why he attached so much importance to a correct understanding

of

the

foreign

text, assuming

that

"anyone

who has

understood a book can also choose the words for a translation if he needs one" (Sierba 1974[1939]:307). The fundamental rule led him

to

the

conclusion

that,

theoretically,

bilingual

19 dictionaries

for

translating

the foreign language

from

the

mother

(L2) are unnecessary

and

tongue

(Li )

into

undesirable.

"If need be, one should resort to large foreign explanatory dictionaries with rich phraseology: there one will always find models for one's own foreign utterances." (Sderba 1974 [1939]:307) Sfierba recognized,

however,

most

a poor

students

have

he therefore

saw a need

that

(due to

command

for

what we would now call the active In

the

typology

of

of

special

foreign

methods)

language,

and

dictionaries

of

type. by

Duda

is taken into

1 source lg.

dictionary

the

teaching

translation

dictionaries

Sfierba's explanatory dictionary

bad

active translation diet. I native passive translation diet. I foreign explanatory diet. I foreign

&

al.

(1986:5f),

account:

target lg.

purpose

foreign native native

text production text production text reception

Table 2 It should be noted the

same

sense

that production

as

in

Table

foreign language; here, any (in

either

language)

translation

always

production

in

Table

not

2 is

dictionaries utterances

in

mother tongue

quite the

In

are

(b) (c) (d)

terms

apply

production. in

respect,

intended

foreign

the

reception

convincing.

to

aid

to

explicitly "four

user

to

in

translating

mentioned

functions must therefore

text

course,

language

fails

by

in the

classification

it

the

without

Of

one

the

Moreover,

language

include all of the classical (a)

called this

- an activity

list of dictionary

There,

are not used

activity which results in a new

is

involves

another. which

1.

and reception

and in

mention

producing from

the

Sfierba.

The

be revised

so as

to

skills":

understanding a text in the foreign language without translation into the mother tongue (reception in L2) translating a text from the foreign language (L2) into the mother tongue (reception in L2, production in LI) creating a text in the foreign language without a model in the mother tongue (production in L2) translating a text from the mother tongue into the foreign language (reception in LI, production in L2) Table 3

20 2.

Multifunctional Dictionaries

A dictionary can serve more than one of the four functions shown in Table 3. The explanatory bilingual dictionary and

the

passive

easiest

to

standing"

translation

combine. and

dictionary

Hausmann

(1977:145f)

"translating

(b))

describes

the

from

(function (a))

(function foreign

are

"underlanguage"

(Herübersetzung) as the two functions of a "reading dictionary" (Lesewörterbuch, a somewhat

misleading

term), and

most

L2-Li

dictionaries exhibit features of both types (cf. also Duda & al. 1986:23).

Just

satisfactorily

as

a

explain

monolingual

meanings

by

dictionary

means

of

can

synonyms,

often it

is

perfectly acceptable to resort to translation equivalents in an explanatory

bilingual

dictionary.

In

Sfierba's view,

they

are

suitable "in all cases where this simplifies the explanation and is in no way detrimental to a full understanding of nature

of

the

often uses

foreign

them

in

word"

his explanatory

(1958). Similarly, passive supplement

(1974[1940]:301) , and

translation

Russian-German

translation

equivalents

dictionary

dictionaries

by

the true Bielfeldt frequently

paraphrases

or

other

comments in order to disambiguate them. This may be necessary, for instance, if the Lz equivalent has a wider meaning than the Li

item

(e.g.

mormor

-

(maternal) grandmother in

English dictionary) or if familiar with

it

(e.g.

the user

cannot be

cariün - gelding

a Swedish-

expected

to be

(horse) in an Irish-

English dictionary). We

should

text and

remember,

translating

however,

that understanding

it into the mother

a

foreign

tongue are different

tasks - a dictionary which serves one of these functions does not automatically serve the other as well. Hausmann may be right that a series of equivalents should "in principle" clarify the meaning of a word actual

dictionaries

with

suffice to

(1977:56), but when one looks at

such

mysterious

translations

as

"passable; omissible"

(for Lithuanian praleidiiamas) , Sfierba's

scepticism

to

understanding

appears is

a

be

necessary

more

realistic.

precondition

Obviously, for

correct

translation. On the other hand, it is not sufficient, äöerba's

21

belief that one can think of a suitable equivalent once one has understood the foreign word was rather naive: If a dictionary explains the Russian word oceski as "what remains when you comb or hackle", a speaker of English will understand what it means but will he know that the English equivalent is combings? A good passive

translation

explanatory dictionary

as

dictionary

well,

can

be

while

will

the

therefore

usefulness

substantially

of

enhanced

have an

if

it

to

be

explanatory also

offers

suitable translation equivalents. Some

lexicographers

have

suggested

that

the

relationship

between the tasks (c) and (d) in Table 3 (free production in Lz and translation from Li into L2) is parallel to that between (a) and

(b) . Hausmann

(1977:145f)

"writing dictionary"

ascribes

both

functions

{Schreibwdrterbuch), Lotzsch & al.

to

a

(1983-

84:VII) claim that "active" dictionaries can be helpful in using the foreign language or (bzw.) translating into it. It should be clear, however, that if we really try to "think in the foreign language", we typically consult a dictionary in order to check our assumptions about the spelling, pronunciation and/or meaning of an hz item that comes to mind or to find out more about its grammatical properties,

collocations

etc. If, for instance, a

speaker of German wants to employ the English word different but is not

sure

dictionary

which of

information

preposition

the

active

under

the

comes

type,

after

which

German

it,

a

might

translation provide

entries

the

verschieden,

unterschiedlich or anders, would obviously be less suitable than a

dictionary

with

English

lemmas

be

it

a

monolingual

(learner's) dictionary or a bilingual dictionary with English as the source language. It

is

thus

not

the

active

dictionary

(or explanatory

production

in

Lz

morphological, language

items.

if

it

syntactic

but

the

dictionary) provides

the

information

Interestingly,

the

required etc.

is

translation used

for

phonological,

about

planned

German dictionary described in Duda & al. such information, which

passive

which can be

the

passive

source Russian-

(1986) is to contain

quite superfluous

for

the primary

"passive" function. The only exception seems to be information on inflection:

The

authors

assume that

a user

who

finds

an

22

inflected

Russian

corresponding

word

citation

in

a text

form;

if

can normally

determine

the

the dictionary

indicates

the

inflection pattern, he can check his assumptions (cf. Duda & al. 1986:53). A much better method would be to explain how to derive citation

forms

1989:736f)

from

and

to

others

enter

the

paradigm

cross-references

lemma for all forms

to which

Principal

parts

paradigm

therefore

not

or

in

required

in

to

(cf.

the

Mugdan

appropriate

the general rules do not apply. numbers a

after

genuine

the

lemmas

passive

are

translation

dictionary. Although Duda & al. (1986:24) assume that the users of their dictionary will have questions about the pronunciation, the

grammatical

properties

or

the

collocations

of

certain

Russian words, they fail to notice that such questions are most likely

to

without relevant

arise

in

a German

the

model

information

course

of

- they with

producing

justify

their

the

a

Russian

inclusion

"descriptive

text

of

the

intentions"

(1986:23), whatever that may mean. We have seen that an L2-Li dictionary can help speakers of Li with three of the four tasks, (a) - (c) . For speakers of L2 , a dictionary with L12 as the source language and Li as the target language

could

mother tongue

only into

active translation

serve

function

the foreign dictionary

(d) , translation

from

the

language. A combination of an

for one speech community with a

passive/explanatory dictionary for the other may at first sight appear to be a step in the wrong direction.

But upon

closer

examination, it turns out that most of the demands the two kinds of dictionaries

should

meet

(cf.

Kromann/Riiber/Rosbach

are compatible with each other. For instance, both and

the

equivalents

can

be

accompanied

morphological data - the problem

of

by

space

1984)

the lemmas

phonological

limitations

or

can be

overcome if the information is restricted to irregularities (the regular

cases

being

covered

by

a

general

statement

in

the

introduction or a grammatical sketch). Occasionally, repetition if

there

seems

to

be

a

danger

of

unnecessary

features of active and passive dictionaries are

combined. Thus, meaning discrimination in the source language is essential in an active dictionary, as in the following EnglishSpanish example from Kromann/Riiber/Rosbach

(1984:201):

23 race

(contest of speed) carrera; (subdivision of mankind) raza.

Since carrera and

and raza

'split, crack',

also have other meanings

respectively)

(1984:201)

(pugna de velocidad) carrera; (subdivisión del género humano)

A combination of the two entries

pace"

they must be disambiguated

a passive dictionary; Kromann/Riiber/Rosbach race

('running

in

suggest:

raza.

(with the explanation after the

equivalent, which seems to be more appropriate) would result in: race

Here,

(contest of speed) carrera (pugna de velocidad); (subdivision of mankind) raza (subdivisión género humano).

the disambiguating

repeat

the

"contest

gloss

information

of

speed"

"pugna de

provided

(although

by

velocidad"

the

it serves

meaning

appears

"subdivisión

del

the entry contained

género

several

a different

humano").

It

to

discriminator purpose

stands in an implicit opposition to "paso del que corre" than

del

could

be

and

rather

omitted

equivalents which disambiguate

if

each

other. In

other

active passive little

the

problems whether

one

the

the

target

either.

For

or

the

Sfierba's

other

bilingual

language

German,

and

try

that

features

(such

Bambara-French),

therefore many

pairs

of

to

active

to and

be

metalanguage

if

it

they

are the

and

should

are

are

such as

a

any

matter

explained

an

of

pose

labels

available, more

to

of

that

not

seems

subject-field

very based

in

both

English

etc.). recent

suggest

work

that,

on

ideally,

the a

for one speech community - but for as

Lithuanian-English,

this will

develop

claim

language

dictionaries

dictionary should be designed many

the

language

French phys[ique]

considerations

of

that

source

example, of

if internationalisms German Phys[ik],

physfics], typology

be

abbreviations

language

(especially

principle

must

dictionary

serious on

cases,

dictionary

not be

a reference

work

"universal"

but

passive

Serbocroat-

feasible. which

-

One

should

unlike

inconsistently

dictionaries

for the widest possible variety of user needs.

-

actually

the mix

caters

24 3.

References

Bielfeldt H[ans] H[olm] (1958): Berlin [DDR]: Akademie-Verlag

Russisch-Deutsches

Wörterbuch.

Duda, Walter / Frenzel, Maria / Wöller, Egon / Zimmermann, Tatjana (1986): Zu einer Theorie der zweisprachigen Überlegungen zu einem neuen russisch-deutschen Lexikographie: Wörterbuch. Berlin [DDR]: Akademie der Wissenschaften der DDR, Zentralinstitut für Sprachwissenschaft (Linguistische Studien, Reihe A: Arbeitsberichte 142) Hausmann, Franz Josef (1977): Einführung in die neufranzösischen Wörterbücher. Tübingen: Niemeyer Arbeitshefte 19)

Benutzung der (Romanistische

Kromann, Hans-Peder / Riiber, Theis / Rosbach, Poul (1984): "Überlegungen zu Grundfragen der zweisprachigen Lexikographie." In: Wiegand, Herbert Ernst (ed.), Studien zur neuhochdeutschen Lexikographie V. Hildesheim, Zürich, New York: Olms (= Germani3-6/84), 159-238 stische Linguistik Lötzsch, Ronald and collective (1983-84) : Deutsch-Russisches Wörterbuch. 3 vols. Berlin [DDR]: Akademie-Verlag Mugdan, Joachim (1989): "Grundzüge der Konzeption einer Wörterbuchgrammatik". In: Hausmann, Franz Josef / Reichmann, Oskar / Wiegand, Herbert Ernst / Zgusta, Ladislav (eds.), Wörterbücher - Dictionaries - Dictionnaires : Ein internationales Handbuch zur Lexikographie - An International Encyclopedia of Lexicography - Encyclopédie internationale de lexicographie. Vol. I. Berlin, New York: de Gruyter (Handbücher zur Sprach- und Kommunikationswissenschaft 5.1), 732-749 Oiegov, Sergej Ivanovifi (1952): Slovar' russkogo jazyka. 2nd edition. Moskva: Gosudarstvennoe izdatel'stvo inostrannych i nacional'nych slovarej Sfierba, Lev Vladimirovifi dejatel 'nost ', ed. L. R. Nauka

(1974): Jazykovaja sistema i reäevaja Zinder / M. I. Matusevifi. Leningrad:

Sfierba, Lev Vladimirovifi (1974 [1939]) : "Predislovie [k Russkofrancuzskomu slovarju]." In Sfierba (1974:304-312) [originally in: Sierba, Lev Vladimirovifi / Matusevifi, Margarita Ivanovna (1939): Russko-francuzskij slovar', 2nd ed. Moskva] Sfierba, Lev Vladimirovifi (1974[1940]): "Opyt obàfiej teorii leksikografii." In Sfierba (1974:265-304) [originally in: Izvestija Akademii Nauk SSSR, Otdelenie literatury i jazyka 1940.3, 89-117] Wiegand, Herbert Ernst (1985): "Eine neue Auffassung der sog. lexikographischen Definition". In: Hyldgaard-Jensen, Karl / Zettersten, Arne (eds.), Symposium on Lexicography II. Proceedings of the Second International Symposium on Lexicography May 16-17, 1984 at the University of Copenhagen. Tübingen: Niemeyer (Lexicographica Series Maior 5), 15-100

Hans Kristian Mikkelsen

"WHAT DID SCERBA ACTUALLY MEAN BY "ACTIVE" AND "PASSIVE" DICTIONARIES?"

L

Introduction

In the present paper I intend to draw attention to the use of the terms "active" and "passive" in connection with dictionaries. In article 285 (= Kromann/Riiber/Rosbach in print) in the not yet published volume two of the International Encyclopedia of Lexicography, active and passive dictionaries are defined in the following way: "By an active dictionary, then, we understand a monofunctional L1>L2 dictionary; by a passive dictionary, a monofunctional L2>L1 dictionary." - where "LI" stands for native language, "L2" for foreign language, and "monofunctional" for the aiming of the dictionary at the speakers of only one of the languages in question. In short, active dictionaries are used to translate from the native to the foreign language, while passive ones serve the opposite function, to translate from foreign to native language. In establishing the active/passive typology as a central concept of bilingual lexicography the authors rely on Smolik 1969, but the mere idea of distinguishing between LI and L2 stems from 1936 and the Russian linguist Lev Vladimirovic Scerba: "Sfcerba, unlike previous lexicographers, thus placed crucial emphasis on the fact that a dictionary user has native-language competence. Out of these ideas have grown more elaborated proposals for a typology of translation dictionaries, operating with four dictionaries per languagepair." (Kromann/Riiber/Rosbach in print). The authors correctly state that ¿cerba did not have the same kind of L2>L1 dictionary in mind as they do. One might further add that §6erba did not mean the same thing by "active" and

"passive" as do

Kromann/Riiber/Rosbach. What I intend to describe in this paper concerns partly the substance of Scerba's bilingual lexicographic theory as expressed in the question "What is an L1>L2, resp. L2>L1, dictionary in Scerba's conception?", partly the terminological question "What did S£erba mean by "active" and "passive"?". It is already indicated by the two different questions that it would be mistaken to identify dictionaries operating between an LI and an L2, i.e. monofunctional bilingual dictionaries, with active and passive dictionaries. In order to answer the two questions one has to embrace the broader

26

historical and linguistic background of Scerba's ideas. Finally, it would be appropriate to try to ask: "Have Scerba's thoughts about bilingual lexicography survived, and if yes, in what form did they survive?" In this respect it is natural to look at trends in modern Eastern and Western lexicography. 2.

Scerba's thoughts about bilingual dictionaries

In the twenties and thirties Scerba was a member of a special lexicographic commission within the Academy of Science, and during this work he himself compiled, or as he preferred to call it, "created" part of the articles of the letter "i" (namely, from "i" to "idealizirovat'sja") in a big, never finished, monolingual Russian dictionary, which had been initiated towards the end of the 19th century. In 1936 Scerba was the general editor of a completely new Russian-French dictionary. New editions, based on the second edition from 1939 of this dictionary, are still being published. t

On the basis of this practical experience Scerba worked out some theoretical reflections which a lot of people consider the beginning of theoretical lexicography. These theoretical conclusions were originally published as the preface to the Russian-French dictionary and as an article in a publication from the department of literature and language under the Soviet Academy of Sciences. The latter was meant as the first in a series of five studies in the general theory of lexicography. It contains the typology of dictionaries, while the following were to elaborate on the nature of the word, its meaning and use, its relations to other words, and, finally, the structuring of semantic, grammatical and stylistic information in the dictionary article. Unfortunately, the first study is also the last: the planned ¥ ones were never written, which was mainly due to the Second World War. Sfcerba died in Moscow in December, 1944. In the following I will be concerned with Scerba's ideas about bilingual lexicography. The main sources are, naturally, the two above mentioned publications, but I will also take other works

Sderba's into consideration, because I am convinced that one has to

include especially his general linguistic and pedagogical views in order to fully understand what his thoughts about dictionaries are all about. t

v

Sierba (Sterba 1940) operates with six binary oppositions in order to establish a typological skeleton. The one which is interesting in connection with bilingual dictionaries is number five, i.e. the distinction between an explanatory and a translation dictionary. It is interesting to notice that the difference between monolingual and bilingual dictionaries is not found typologically important. Generally, one might say that Scerba's opinion of the so-called translation dictionaries is very low, and that goes not only for the evaluation of the existing instances of this type, but also for the concept of translation dictiona-

27

ries as such. It is, therefore, ironical that the bilingual dictionary for which Scerba himself was responsible can only be classified and understood as a translation dictionary. The translation dictionary is characterized by Scerba as a malum necessarium, a necessary evil, as long as translation from native to foreign language is in question, and until the ideal onomasiologically based dictionary can be realised. Recognizing that the fulfilment of a true onomasiological dictionary was totally unrealistic at the time when he presented his thoughts, Sterba instead decided to get the best out of the translation dictionary. For translation from L2 to LI he proposed a mixture of an explanatory and a translation dictionary to which I will return. But let us first look at the concept for an L1>L2 dictionary. Sierba's main complaint about L1>L2 dictionaries is that they do not give the user the help he needs in order to produce translations which will turn out to be, not only understandable, but also non-ridiculous. The problem is that the lexicographers have not duly considered the user's L2 or lack of L2 competence. What is needed, then, for a Russian who wants to translate into French, is a dictionary which can lead her or him safely to an equivalent which will fit into the given context. Sterba therefore formulates five principles whose essence can be rendered as follows: 1) Provide a translation, not an explanation, that will, in the appropriate grammatical form, fit into a correct French sentence which has been translated from a Russian sentence. Of all the possible candidates choose the one which fits into most of the Russian contexts. If no general equivalent is found, make sure that the intended user has enough information to judge which one will serve him best. 2) Reject the translations which are "too French" and metaphorical, and take only the simplest ones in order not to let the user seem ridiculous. 3) Throw away all approximate translations and synonyms - again in order not to make anecdotal translations possible. 4) If no precise equivalent is available, give the approximate ones together with the corresponding explanations. If not even an approximate equivalent can be found, leave the lemma as untranslatable, followed by an explanation in brackets. Under certain circumstances bring a translation of whole contexts. 5) Provide the necessary grammatical information, so that the user can produce the correct morphological and syntactic forms. It is evident from these methodological principles that Scerba has a very modest opinion of the use of an L1>L2 dictionary. The main advantage of L1>L2 dictionary, compiled

28

on this basis, is that it will help an LI user, who has very little knowledge about L2, to make correct, simple translations. The "tuning" of the dictionary to an LI (Russian) user on a rather low level of acquaintance with L2 (French) entails that an L2 user, i.e. a Frenchman, will not be well suited with this dictionary. The reason for this lies partly in the microstructure - as can be seen from the five principles mentioned - partly in the macrostructure. With regard to the macrostructure the selection of lemmata in an L1>L2 dictionary does not coincide with the selection in an L2>L1 dictionary. The L1>L2 dictionary must consider what can be translated into an active vocabulary, typical of an L2 user. The L2>L1 dictionary, on the other hand, must include the passive vocabulary which is typical for the L2 user. In other words, a Russian-French dictionary for a Russian user must include only those Russian words which the user can translate into a French active vocabulary, while a Russian-French dictionary for a French user must concentrate on the Russian passive vocabulary. This means that the French user cannot expect the RussianFrench dictionary for a Russian user to contain information needed to read e.g. classical Russian literature. Further, it means that the lexicographer cannot simply make a left side out of the right side from existing dictionaries which operate in the opposite direction. The practice in Russian-French pre-Scerbian lexicography built upon this method, due to the fact that the tradition in Russian-French/French-Russian lexicography began with the collection of Russian equivalents to French explanations in French monolingual dictionaries. As far as the microstructure is concerned, one can foresee the problems an L2 user may have interpreting the usage information, provided for the benefit of an LI user, the most evident feature being the formulation of metalinguistic information in the LI. I.e., it is not appropriate for a Frenchman to have this type of information in Russian. Apart from this v

obvious practical hindrance, there is another problem, which in Scerba's opinion is even worse, namely that a translation dictionary never conveys the understanding of any of the languages involved. And here we come to the problem of L2>L1 dictionaries. Scerba's main objection against translation dictionaries is that they do not provide information about the language system, but rather pretend isomorphism, or as Scerba puts it "adequacy", between the conceptual systems of the languages involved. This criticism can be ignored only when L1>L2 translation of the most primitive kind is in question, and this purpose constitutes the raison d'être of such dictionaries. But, when it comes to L2>L1 translation, according to Scerba, one can no longer leave out of account the obligation of the dictionary to convey real information about the foreign language. The problem is that the LI equivalents in such a translation dictionary have their own meanings

29

and nuances which may or may not coincide with the meaning potential of the L2 lemma. Therefore, one can only use such an L2>L1 dictionary for the purposes of reading in and translating from L2, because it serves to exhibit potential LI equivalents. In other words, it lets the user make a qualified guess about the contextually correct, actual equivalent (It may be recalled from the reflections about the L1>L2 dictionary that actual equivalence is exactly what is aimed at in that kind of dictionary, an intention which is only made realistic by the restricted selection and the principle of presenting just the most generally applicable equivalents.) It is essential for ¿cerba that the L2>L1 dictionary also gives information which can be used the other way round, that is, when the user needs to translate from LI to L2. Normally information about the paradigmatic and syntagmatic characteristics of a given lexical entity can be found in the typical monolingual dictionaries. What is needed is then a fusion of the explanatory and the translation dictionaries. In principle there is nothing that speaks against simply translating monolingual dictionaries of the L2 into the LI, thus taking over the monolingually relevant lexicographic principles. This means, among other things, that the semantic structuring which is relevant for the monolingual description is automatically found to be the right one also in the bilingual L2>L1 dictionary. The rule is, consequently, that the definitions or explanations to the L2 lemmata are accepted without ceremony. Exceptions are possible: if an LI equivalent can be found which does not disturb the full comprehension of the L2 word's true nature it can replace the explanation. Now, it is clear why Sfcerba became famous for his statement about the necessity for four user-profiled bilingual dictionaries for each pair of languages. Only, I find that this statement is often exaggerated in a way that, I am sure, Scerba would not have agreed with. Without discussing here to what extent Sderba's conception of bilingual dictionaries coincides with what is nowadays referred to as "active" and "passive" dictionaries, I want to stress that Scerba only stated the need for L2>L1 dictionaries of the above mentioned hybrid nature. The L1>L2 dictionary - of which Scerba's own Russian-French dictionary is an instance - can on the other hand, at least theoretically, be dispensed with as far as the ideal mastering of the L2 goes not via the LI or any other language except for the L2 itself. I will return to the reason for this the following sections. i

Evaluation of Scerba's concept of bilingual dictionaries

Already in 1936, i.e. the year the first edition of Scerba's Russian-French dictionary was published, Lucien Tesniere, the famous French slavist and russist, wrote a short review of this small precious stone, as he called it, hoping hereby not to let it disappear in the vast anonymous mass of commercial, industrially produced, dictionaries. Tesnifcre's

30

manuscript was published posthumously in 1958 and in a Russian translation (Tesnifcre 1958). Tesnifcre draws attention to the thesis requring four dictionaries for each pair of languages and calls it revolutionary. He compares the inclusion of the user aspect to the contemporary development in the Russian theatre, where the audience was being included as a participant on the same level as the author and the actors. Transferred to lexicography the author corresponds to the dictionary, the actors to the words included in the dictionary and the audience to the user. It goes without saying that there is a direct line to Tesniere's own syntactic theory of actants. He also uses another metaphor to explain the difference between an L1>L2 and an L2>L1 dictionary. The dictionary is like a ticket "there-and-back": the Frenchman needs a ticket from Paris to Moscow and back from Moscow to Paris, whereas the Russian needs a return ticket to Paris. Just as it is not necessary for a travelling Frenchman to have both a "there" and a "back" ticket in the same direction, e.g. Paris-Moscow, so it is also unnecessary to have a dictionary which helps both the French and the Russian traveller. Tesniere welcomes and wholly accepts Scerba's concept. It is interesting to notice that new editions of Scerba's Russian-French dictionary seems to be useful also to Frenchmen, although not intended to be so. As a part of the grammatical back matter an overview of Russian morphology with commentaries in French is now published together with the the other material, which is primarily intended for the Russian user. It seems clear that apart from the mere requirement for the user profiling of bilingual dictionaries there are some quite serious differences from the way bilingual lexicography is practised today. In the following I will point out these differences, looking at Scerba's bilingual concept from the angle of the modern state of research. In addition to exposing what these differences consist of, it will be natural to ask why it is that Scerba's lexicographic viewpoints are as they are. 3.1.

In the context of the present state of research

The demand for more than the classical two bilingual dictionaries for each pair of languages did not stop with Scerba's four, but has, as it is well known, been extended to the requirement of both six and eight dictionaries, or at least: dictionary functions. On the basis of the discussion of these problems in Hausmann

1977, Karl 1982,

Duda/Frenzel/Woller/Zimmermann 1986 and Duda 1986,1 think it is fair to conclude that while translation from LI to L2 always tends to aim at the providing of actual equivalents, i.e. equivalents which can be put directly into the context, the translation from L2

31 to L I seems to cover three different aspects. Correspondingly, an L2>L1 dictionary may be intended to cope with 1) potential equivalence, which is needed when one wants to understand a text in an L2, 2) actual equivalence, which is what one needs in order to produce a translation in the L I , and finally, 3), comprehension of the system of L2. The ultimate position is to require one dictionary for each of these four functions, and adopted to a native user from each of the languages included. That makes up eight dictionaries. The more moderate claim for six dictionaries draws upon a two-fold distinction in the translation process. So, e.g., the GDR team behind the new Russian-German dictionary operates with a distinction between production- and reception-oriented translation from L2 to L I , where the latter is the objective of their dictionary. The development in theoretical bilingual lexicography demonstrates that the most successful of Scerba's bilingual dictionary types is the one that copes with L1>L2 translation. His L2>L1 concept has not survived, at least not in the intended form. In a way this may v

seem ironical, considering how low Scerba estimated the pedagogical value of translation dictionaries. By the way, Scerba used the term "translation dictionary" only in the meaning L1>L2 dictionary. His ideal L2>L1 dictionary is, as already said, primarily an explanatory dictionary. The requirements concerning the macro- and microstructure of L1>L2 dictionaries, as they are presented e.g. by Kromann/Riiber/Rosbach 1984 and in print, are very much the same as the principles formulated in the Russian-French dictionary. An L1>L2, or in Kromann/Riiber/Rosbach's terms an active dictionary's microstructure is different from an L2>L1, i.e. a passive, dictionary because of the meaning discriminating and compensating comments, the idiosyncratic phrases and a maximum of grammatical information on the equivalent side. Moreover, the macrostructure, i.e. the selection of lemmata, is rather restricted with regard to regionalisms, special vocabulary, orthographical and morphological variants. On the other hand, what is considered the ideal L2>L1 dictionary today, has not got very much in common with Scerba's translated explanatory dictionary. According to Kromann/Riiber/Rosbach the L2>L1 dictionary in opposition to an L1>L2 dictionary is characterized by an undifferentiated lining of equivalents, by omission of transparent idiosyncratic phrases, and by non-transparent morphosyntactic information. All in all, this concept of L2>L1 dictionaries corresponds more directly to the L1>L2 dictionary. Both concepts are developed from the kind of translation dictionaries where no attention is paid to the user. From a Scerbian point of view the concept of bilingual dictionaries has been rather much narrowed down. Although the characteristics of the modern L2>L1 translation dictionary

32

are logically connected to the L1>L2 dictionary, and one would find it natural if the man behind the L1>L2 dictionary would at the same time have invented the L2>L1 translation dictionary, there is no trace whatsoever of such reflections in Scerba's lexicographic theory. Instead of finding inspiration in the L1>L2 concept, thus taking the first steps towards an autonomous bilingual lexicographic theory, Scerba preferred to believe in the monolingual lexicographic principles as better suited for ail ideal L2>L1 dictionary. One of the few, or perhaps the only, instance of an L2>L1 dictionary, worked out withV

out knowing it in accordance with Scerba's principles, is the Russian-German dictionary from 1958, edited by H.H. Bielfeldt. It was explicitly intended for German users, but has undergone severe criticism, especially for leaving out German equivalents, for not separating explanation and equivalent, for presenting examples and derived words without German translations. These points may be said to concern the concrete manifestation of the Scerbian L2>L1 concept, but there is another criticism which seems to invalidate the concept itself. According to Duda 1986 it is questionable, whether this kind of dictionary enables the user to go from the comprehension of the L2 to the translation into LI. The solution, as Duda sees it, is to change the priorities of explanation and equivalent, so that the translation aspect is considered more important than the comprehension aspect. Nevertheless, Scerba's concept of L2>L1 dictionaries has not completely disappeared. In 1988 a Soviet lexicographer, V.K. Scerbin, claimed that Scerba's criticism of bilingual dictionaries is just as relevant today as it was in 1940. Scerbin repeats Scerba's statement that an L2>L1 dictionary must provide knowledge about the foreign language and not only give hints for the finding of contextually bound equivalents. Scerbin refers to V.G. Gak and S.S.-D. Kim for raising the same criticism, and they all agree that an L2>L1 dictionary must at the same time serve as the basis for production in the L2. It is, then, possible to conclude that the bilingual apple of Scerba's eye, i.e. the L2>L1 concept, has had little, if any, impact on today's theory and practice in ordinary bilingual lexicography, while his stepchild, the L1>L2 concept, has gained overwhelming success. The truth is, however, not without modifications. In chapter 5 I will try to show how Scerba's L2>L1 concept has survived in parts. 3.2.

In the context of Scerba's linguistics

Lev Vladimirovic Scerba continued the tradition of the so-called Kazan' School, whose initiator was Baudouin de Courtenay. The school is today known as the Leningrad

33 School, as opposed to the Moscow School, which stems back from Filip Fedorovic Fortunatov. One of the characteristics of the Leningrad School has always been the focusing on the human being in connection with language. In the work of Scerba this guideline can be followed everywhere, e.g. in his general language theory and in his pedagogical theory and practise, as well as in every application of theoretical principles. I will here stick to Scerba's bilingual lexicographic principles in order to try to make clear how such superordinate viewpoints constantly guide Scerba. One might ask why Scerba is so sceptical towards the mere concept of translation dictionaries that he completely rejects it as a principle for L2>L1 dictionaries, while he emphasizes the limitations of the L1>L2 translation dictionaries. Apart from the limitations resulting from the specific user profiling there are also certain limitations for the intended LI user (see above). The reason for this negative attitude can be sought in the methods and reflections about the purpose of second language learning, defended by Scerba. And these viewpoints are again closely connected to his thoughts about bilingualism. As for bilingualism, defined as the ability of a group of people to speak two languages, Scerba distinguishes between a "pure" and a "mixed" type. In the pure bilingualism the speaker has at his disposal two different language systems which he does not mix, whereas the mixed type is precisely characterized by mixing together the two systems. In a pure bilingual relation the differences between the two language systems is maintained; this is illustrated by among other things the introduction of loan words from one language to the other, in order to compensate for the lacking equivalence. In contrast to this, the mixed bilingualism tends to eliminate the differences between the two languages in such a way that it tends to make one system out of the two. (The impacts of mixed bilingualism finds its extreme in the Sorbian situation. The Slavic language Sorbian, which served as the source for Scerba's doctoral dissertation in 1915, was at that time a forbidden language spoken by a minority and completely surrounded by German. All the Sorbian speakers are bilingual German-Sorbian, and this has developed into a situation, where the only thing that is left of the original Sorbian language is the expression side, while the content has been wholly identified with the German. That goes not only in the lexicon, but also for the grammar, where, contrary to most Slavic languages, an article has been introduced into the noun system and a perfect tense into the verb system. In consequence, Scerba treats the Sorbian case as pseudobilingualism; there is in fact only one language in question, but a language with two expressions, or "terminals". (Scerba 1925))

34

It is evident that the aim of foreign language teaching cannot be a mixed bilingualism, with the consequence that the Russian pupil transfers his own language system into the other, e.g. French. It therefore seems that the so-called "natural" or "direct" method, which focuses on unconscious and intuitive language learning, must be the pedagogically best suited way to a foreign language. As a matter a fact this is exactly the attitude against which Scerba reacts. There are two reasons for this (Scerba 1934): one concerned with the circumstances under which children, and human beings generally, learn foreign languages, and one concerned with the pedagogically superior goals of general education: Conscious learning is the only one practically possible in the Soviet society - in contrast to pre-revolutionary Russia where it was possible for the nobility to virtually isolate the children together with their foreign governesses. Consequently, one might as well make a virtue of necessity. Scerba sees the virtue of conscious foreign language learning in its bearing for the native language comprehension. The device is that the best way to learn ones own language is to learn a foreign language, and further, to emancipate the thought from its language bonds. If the L2 is important for the L I , the opposite can also be stated, i.e. knowledge of the LI can be positively exploited in the L2 teaching. Scerba speaks in favour of a widely application of contrastive language descriptions. (His famous French phonetics (Scerba 1937) is built on this principle). On the other hand, the contrastive method only goes for L2 learning at the beginner's level, because in the end it will lead to a mixed bilingualism due to numerous transfers from L I . Therefore, in order to refine the L2 skills the pupil must begin to leam thinking in the L2. This digression into other spheres of Scerba's activities may shed light upon his concept of bilingual dictionaries. One might tentatively divide the users into two groups: the beginners and the advanced learners. The bilingual dictionaries, both the L1>L2 and the L2>L1 ones, are exclusively intended for the beginners. Only they need to leam the L2 through the L I . In translating from LI to L2 there is no other way to go for the beginner than to translate from L I . The drawback of this way of producing L2 text is that it easily leads to false transfers from L I . Therefore, the L1>L2 dictionary must carefully follow the principles of a restricted macrostructure and a microstructure which contains only the most general equivalents, and if there are more than one, the equivalents must be clearly differentiated from the others. It is obvious why this dictionary can only cope with the most simple expressions of the L2. The advanced learner, on the other hand, must get rid of the L I as an

35 auxiliary tool for L2 learning, and instead start to think in the L2. This also means that he will no longer need the "necessary evil" of the translation dictionary. The argument for the L2>L1 explanatory dictionary is as follows: Again the beginner is forced to learn the L2 through the LI. But at the same time he can learn to look at his own language, the LI, as a system of its own. In order to do this he must have direct access to the L2 system, and this can only be done by providing the L2 lemma with its paradigmatic and syntagmatic surroundings. Translation is in the respect of language learning of very little importance, for which reason it can more or less be ignored. At a more advanced stage of L2 learning the pupil is better off with a monolingual dictionary. Scerba's understanding of "active" and "passive" After the discussion of Scerba's ideas about bilingual dictionaries, I come to the terminological question about Scerba's understanding of "active" and "passive". Let me be quite straightforward about this: Scerba does not use these terms in combination with the word "dictionary", at least not in the literature I am acquainted with. That does not mean that it is impossible that these terms, which are currently the generally accepted substitutes for the term "a monofunctional L1>L2 dictionary" (= active dictionary) and "a monofunctional L2>L1 dictionary" (= passive dictionary), might originally come from Scerba. As a matter of fact these notions are very important for his general language theory as such, and for its various applications, e.g. in pedagogics. On the other hand, in Scerba's use, active and passive are not confined to bilingual matters. Let me therefore try to explain the meaning of these terms, primarily in the context of Scerba's language theory, but also in its most important pedagogical implications. In 1931 Scerba formulated his own threefold language theory,

to a certain degree

similar to that of Saussure. But, contrary to Saussure's bipartite distinction between langue and parole, Scerba's theory contains three aspects, namely the system, the activity, and the material. The system consists of grammar and lexicon, the activity of the processes of speaking and understanding, whereas the material consists of texts, i.e. what is actually spoken (/written) or understood. The three aspects are internally ordered: the text presupposes the activity, which again presupposes the system. The system itself is extra-linguistically determined by time and place, i.e. by an actual society. The system is always hidden for the unconscious language user. That is why the system of a native language has to be learned. If we take a look back on Scerba's bilingual dictionaries, we will see that the explanatory L2>L1 dictionary intends exactly to reveal the language system of the L2, and indirectly also of the LI.

36 The terms "active" and "passive" are only used in connection with the activity aspect of language. They cover the two already mentioned processes: speaking and understanding, respectively. It seems all right to call Scerba's L1>L2 dictionary "active", because its function is to make possible a process of speaking. But it would be mistaken to assign the label "passive" to the L2>L1 dictionary, because what it primarily does is not to facilitate an understanding activity of the L2 text, but rather to provide insights into the language systems. It is interesting to notice that the thinking skills in Scerba's conception are only directly accessible through the language system. That is also why it is so important to stress the positive effect of L2 learning with regard to system learning, and thereby to the emancipation of the thought from the language. Active and passive language skills are relevant for both native and foreign language communication. It is therefore not surprising that these terms are used in connection with monolingual, as well as bilingual questions. Scerba speaks explicitly about an active and a passive grammar. A passive grammar is one which helps to understand, i.e. to get from the expression to the content, whereas an active grammar operates the other way round: from the content to the expression. In addition to the Russian-French dictionary Scerba wrote a short active grammar, which serves exactly the same goal as the dictionary itself, namely to get from content conceived in Russian to the French expression. (The overview of Russian morphology, which I mentioned earlier, has the characteristics of a passive grammar and is intended for the French user. It was written by I.M. Pul'kina and was later developed into her famous Short Reference Grammar of Russian.) It can be concluded that Scerba uses the terms "active" and "passive" much the same way as the terms "onomasiological" and "semasiological" are used. In fact, Scerba might well have applied the distinction for one of his dictionary types. I have in mind the fourth opposition, the one between an ordinary (explanatory or translation) and an ideological dictionary, i.e. the distinction between an alphabetically (semasiologically) and a notionally (onomasiologically) structured dictionary. These dictionaries are the ones to take over when the learner leaves the beginner's level. Finally, the active/passive concept is used in combination with "vocabulary". It is essentially the same meaning: an "active vocabulary" is the set of lexical items one uses as output when speaking, whereas the "passive vocabulary" is the much bigger set of lexical items which serves for understanding. I have already mentioned the impact of these notions on the macrostructure of the two bilingual dictionaries (cf. chapter 2 above).

37

L

A general evaluation of Scerba's lexicographic views

Finally, I want to point at some lines of development not directly connected with the theme of this paper, but nevertheless important in a general lexicographic aspect. I stated above that Scerba's L2>L1 concept seems to have been shipwrecked, with the consequence that the pre-Scerbian ideal of a translation dictionary has been renewed and, of course, improved by the user-friendly tuning. But I think the core of the L2>L1 concept has survived in the disguise of the so-called learner's dictionaries. This field is without doubt the most exciting in contemporary Soviet lexicography. It has developed its own theoretical device which is presented by its leading figure V.V. Morkovkin in Morkovkin 1986 and Morkovkin 1986a. One of the main points concerns the adjustment to different user types. The typical situation is that the learner's dictionaries are monolingual, but oriented towards students for whom the given language is an L2. In this respect the connection to Scerba's L2>L1 dictionary is quite evident, the only major difference being that the explanatory dictionary is monolingual, while Scerba's concept is bilingual, although this bilinguality is secondary. Another point is that the learner's dictionary focuses on different activities, just as Scerba does. The distinction between active and passive vocabulary is clearly drawn. (Morkovkin 1986a: 112). There is a growing interest for Scerba's theory in the Soviet Union. It is a pity that most of his written works are only available in the Russian original, because I think there is yet a lot to be learned from Scerba - also for lexicographers. Let me just mention that one of the actual questions in the present discussions, the distribution and definition of grammar and lexicon, has also been solved by Scerba. iL

Conclusion

The main conclusions to be drawn are the following:

- Scerba distinguishes between two quite different bilingual dictionaries of which the one (L1>L2) is for translation, while the other (L2>L1) is primarily explanatory. - In Scerba's opinion bilingual dictionaries are only suited for non-advanced users. They are best substituted by monolingual semasiological (for understanding) and monolingual onomasiological (for speaking) dictionaries. In the long run bilingual dictionaries are harmful, because they provide the basis for mixed bilingualism by pretending interlingual isomorphism.

38 - Scerba's L1>L2 dictionary is still a paragon for this kind of translation dictionaries. - Scerba's L2>L1 dictionary ideal is applied in the lexicography of learner's dictionaries with slight modifications. Its character as an explanatory dictionary is not intended to provide translations, but have an L2-, and also Ll-constituting function. This is the initial step in conscious foreign language learning. It is not, however, considered relevant in modern bilingual lexicography, which has instead chosen to tune the traditional translation dictionary according to the LI user. - Scerba's lexicographic thoughts must be seen together with his pedagogical theory of foreign language learning and with his general language theory. - The terms "active" and "passive" are used frequendy in Scerba's works, but never with dictionary. They can be understood as onomasiological and semasiological.

39 Bibliographv Duda, W. 1986. Ein "aktives" russisch-deutsches Wörterbuch für deutschsprachige Benutzer?. Beiträge zur Lexikographie slawischer Sprachen 1986, Günther, E. (ed.) (= Linguistische Studien. Reihe A. Arbeitsberichte. Akademie der Wissenschaften der DDR. Zentralinstitut für Sprachwissenschaft 147). Berlin, Akademieverlag. 9-15 Duda, W., M. Frenzel, E. Wöller, T. Zimmermann. 1986. Zu einer Theorie der zweisprachigen Lexikographie. Überlegungen zu einen neuen russisch-deutschen Wörterbuch. Linguistische Studien. Reihe A. Arbeitsberichte. Akademie der Wissenschaften der DDR. Zentralinstitut für Sprachwissenschaft 142. Hausmann, F.J. 1977. Einführung in die Benutzung der neufranzösischen Wörterbücher. (Romanistische Arbeitshefte 19) Tübingen, 1977. Karl, I. 1982. Linguistische Probleme der zweisprachigen Lexikographie. Eine Nachlese praktischer Wörterbucharbeit. Linguistische Studien. Reihe A. Arbeitsberichte. Akademie der Wissenschaften der DDR. Zentralinstitut für Sprachwissenschaft 96. Kromann, H.-P., T. Riiber, P. Rosbach. 1984. Überlegungen zu Grundfragen der zweisprachigen Lexikographie. Studien zur neuhochdeutschen Lexikographie V, Wiegand, H.E. (ed.) Hildesheim, New York, Georg Olms Verlag, 159-238. Kromann, H.-P., T. Riiber, P. Rosbach. In print. Principles of bilingual lexicography (art. 285). Wörterbücher, Dictionaries, Dictionnaires. An International Encyclopedia of Lexicography, Second Volume, Hausmann, F.J., O. Reichmann, H.E. Wiegand, L. Zgusta (eds.), Berlin, New York, Walter de Gruyter. Morkovkin, V.V. 1986. O sostojanii i zelatel'nych perspektivach russkoj ucebnoj leksikografii dlja inostrancev (vmesto wedenija). Ucebniki i slovari v sisteme sredstv obucenija russkomu jazyku kak inostrannomu. Sbomik statej, Morkovkin, V.V., L.V. Trusina (eds.), Moskva, Russkij jazyk, 94-102. Morkovkin, V.V. 1986a: O bazovom leksikograficeskom znanii. Ucebniki i slovari v sisteme sredstv obucenija russkomu jazyku kak inostrannomu. Sbomik statej, Morkovkin, V.V., L.V. Trusina (eds.), Moskva, Russkij jazyk, 102-117. Scerba, L.V. 1925. O ponjatii smesenija jazykova. Scerba, L.V., Jazykovaja sistema i recevaja dejatel'nost', Leningrad, Nauka 1974, 60-74. First printed in French: "Sur la notion de mélange des langues", Jafeticeskij sbornik 4, Leningrad 1925,1-19. Scerba, L.V. 1931. O trojakom aspekte jazykovych javlenij i ob éksperimente vjazykoznanii. Izvestija AN SSSR, Otdelenie obscestvennych nauk and Scerba L.V., Jazykovaja sistema i recevaja dejatel'nost', Leningrad, Nauka 1974,24-39. Sèerba, L.V. 1934. O vzaimootnosenijach rodnogo i inostrannogo jazykov. Inostrannyj jazyk v srednej skole 1 and Scerba, L.V., Jazykovaja sistema i recevaja dejatel'nost', Leningrad, Nauka 1974, 338-343. Scerba, L.V. 1937. Fonetika francuzskogo jazyka. Scerba, L.V., Jazykovaja sistema i recevaja dejatel'nost', Leningrad, Nauka 1974, 125-135. Scerba, L.V. 1939. Predislovie ko vtoromu izdaniju. Scerba, L.V., M.I. Matusevic, Russko-francuzskij slovar1, Izd. 2-e, rassir. i pererab, 1939, 6-9 and Scerba L.V., Jazykovaja sistema i recevaja dejatel'nost', Leningrad, Nauka 1974,304-312.

40

Scerba, L.V. 1940. Qpyt obscej teorii leksikografii. Izvestija AN SSSR, serija literatury i jazyka 3 and Scerba, L.V., Jazykovaja sistema i recevaja dejatel'nost', Leningrad, Nauka 1974, 265-304. Scerba, L.V. 1974. Jazykovaja sistema i recevaja dejatel'nost', Leningrad, Nauka. Scerbin, V.K. 1988. K probleme postroenija russko-nacional'nogo tezaurusa. Slovarnye kategorii. Sbornik statej, Karaulov, Ju.N. (ed.), Moskva, Nauka, 1988, 213-216. Smolik, W. 1969. "Aktives" Wörterbuch Deutsch-Russisch. Nachrichten für Sprachmittler 3,11-13. Tesnière, L. 1958. (Ten'er, L.) O russko-francuzskom slovare L.V. Scerby. Voprosy Jazykoznanija 6,41-43.

Ulrich Heid SYNTACTIC INFORMATION IN (MACHINE) TRANSLATION D I C T I O N A R I E S — towards a modular architecture for bilingual dictionaries 1

1. Introduction T h e subtitle of this paper corresponds to a more general question which will be addressed, namely t h a t of modularity and (possibly as a consequence of it) reusability of lexical descriptions in different applications of natural language processing. We will particularly consider machine translation or natural language generation systems which are concerned with more than one language pair. T h e title refers to the specific problem which is addressed, and which, in some sense, is used as an example for t h e more general idea: t h e description of syntactic properties of lexemes, the interaction between these properties and phenomena which one would "naively" classify, in the approach we are going to use, as lexeme-independent (or: grammatical). Finally, t h e question of how to represent the results of this manifold interaction of different phenomena in a bilingual dictionary, on paper or in a machine-readable form, will be discussed. We will take t h e relatively well-understood domain of subcategorization properties of verbs (complement types and complement realization, number of complements, etc.) as an example for the complex descriptions needed for a translation dictionary, be it an "active" dictionary for "human use" or a machine translation dictionary. We will claim that with a modularized dictionary architecture which allows for a combination of source and target language monolingual descriptions, related through language-pair based directional "overlays" which make use of predicate-argument structures, we achieve descriptive clarity and reusability. In our view, this is an advantage with respect to current approaches to the construction of bilingual dictionaries, both in lexicography and in machine translation. This claim will be illustrated with some examples from French «-» German translation. ' T h e work discussed in the present paper has partly been carried out within the POLYGLOSS project, funded by the German Minister of Research and Technology (BMFT: Bundesminister fur Forschung und Technologie), under grant no. 08 B 3116 3.

42 T h e discussion of alternative lexicographic approaches will be limited to the most well-known theories of bilingual dictionaries, namely that of [KROMANN/RIIBER/ROSBACH 1984] etc. which we will call the "directional approach", and t h a t of [VAN STERKENBURG/MARTIN/AL 1982], etc., which will henceforth be called the "concept-based approach". T h e paper starts with a description of some of the problematic data which should be described in a good bilingual dictionary, be it intended for the human translator or writer of texts in a foreign language 2 or for machine translation (section 2.). T h e discussion of these facts will lead to some requirements concerning the descriptive "coverage" of a bilingual dictionary (section 2.3.). A comparison of the "directional approach" and the "concept-based approach" will then show some of their differences with respect to these requirements (section 3.). Even if they may seem, at a first glance, nearly complementary from the lexicographic point of view, I will put forward a proposal for t h e architecture of a machine translation dictionary which combines advantages from both (section 4.). Examples taken from a system under development in P O L Y G L O S S 3 , a multilingual generation project implemented in a typed feature unification formalism, are intended to show how such a combination may be achieved. T h e claim here is t h a t it is possible (1) to minimize redundancy in a machine translation dictionary by building up classes of lexical objects and formulating generalizations for those classes, and (2) to implement the modular conception of bilingual dictionaries as combinations of separate source and target language dictionaries and languagepair based equivalence statements.

2. Problems of syntactic description in bilingual dictionaries 2.1. A preliminary remark: description vs. presentation Metalexicographic discussion is concerned, among others, with the way dictionaries describe, arrange, organize and present partial linguistic information for a (more or less clearly definable) group of users. Metalexicographic work aiming at 2 I agree with [MUGDAN 1990], in this volume, in assuming t h a t the access path to lexical information in a dictionary need not be the same in translation towards a foreign language and in writing in a foreign language. This is, however, more a point concerning the presentation of bilingual information than the lexicographic description itself. Since this paper will concentrate on descriptive rather than on presentational issues, the distinction between both skills may not be indispensable here. 3 For an overview of the project, especially from the point of view of the architecture of the knowledge sources, see [EMELE/HEID/MOMMA/ZAJAC 1990B].

43

a critical description of existing dictionaries or at proposals for future ones, has of course to keep track of needs and knowledge of the users; much of the metalexicographic discussion is therefore concerned with the way dictionary authors present the results of their description of linguistic data. This includes devices used by dictionary authors in order to emphasize certain facts, to make them accessible to a user, or last but not least, to avoid redundancy and information regarded as unnecessary with respect to the intended user group. These problems are highly dependent on the user group of the dictionary, but do not necessarily affect the lexicographic data description as such. At least it may be claimed — and we are going to make this claim — that the linguistic and lexicographic description and the way how the descriptive results are presented to the user have to be described separately; this includes the claim that the description of the actual linguistic facts may proceed the same way, irrespective of user groups, even if the facts are presented in different ways, or if different subsets of the facts are presented to different groups of users. This is what "databases":

[MARTIN/AL

1988] mean by "characteristics" of "front-ends" and

"User-orientation is to be defined as a relative characteristic of dictionaries (front-ends), not (necessarily) of the underlying databases from which they are derived" Separating between data description and presentation of descriptive results, both seen as two basic activities in dictionary making, seems therefore to me to be also a good principle of the metalexicographic description of existing dictionaries and of proposals for dictionary architecture. Within the description of dictionaries for machine translation, the presentational requirements are of course less important than in lexicography for "human users", since all descriptions have to be presented in an equally explict form. No a priori knowledge or inferencing from knowledge sources outside the dictionary is possible. There are, however, good reasons to assume a parallelism between dictionaries for "human use" and dictionaries for machine translation with respect to the requirements concerning the data description4.

2.2. Contrastive phenomena in bilingual dictionaries [ K R O M A N N 1989] describes the task of bilingual lexicography as a comparison of two non-isomorphic linguistic systems. This lack of isomorphism between two languages which are "confronted" may occur at any level of the linguistic description 4I

discussed this point in detail in [HEID 1 9 8 8 A ] .

44 of linguistic objects or at any combination of such levels. Some of these levels are constituent-structure, functional or thematic structure or, most strikingly, "lexical semantics". In this paper, we will concentrate on problems of anisomorphism at the level of syntactic description (e. g. constituent structure, grammatical functions, etc.). Problems which arise from this type of lexical differences between two languages manifest themselves in translation by the following fact: even if in both languages there potentially exist structurally isomorphic constructions, in the context of some lexemes, this "evident" way of translating is impossible, and alternative translations have to be sought. The user would like those cases to be explicitly treated in a translation dictionary: he is of course most interested in contrastive information in "critical cases".

2.2.1. "Realization mismatches" The comparison between the grammatical systems of French and German does not lead a priori to evidence which would suggest that the two languages differ with respect to the existence of verbs which can take subject clauses. Verbs used to express, for example, two-place relations with a propositional argument as the first argument, may in both languages take subject clauses: (1) Que tu ne fumes pas, me plaît. (2) Daß Du nicht rauchst, gefällt mir. (3) Daß er zu spät kommt, erstaunt (4) Qu'il vienne trop tard

mich.

m'étonne.

There exists, however, a certain difference between the two languages at the level of the syntactic behaviour of individual verbs: some German verbs allow for a subject clause, whereas their French equivalents do not. The German verb begeistern can take a subject clause, whereas the isomorphic construction with its French equivalent enthousiasmer is unacceptable in written language (code écrit 5 ), and translators thus use nominalizations (9) as translations of clausal complements (6). 5 There may exist, in spoken French, (code parlé), possibilities of building utterances like (5) with extraposition and segmentation.

(5) Que /'experiment

réussisse, ça enthousiasme

le chercheur.

45

(6) Daß der Versuch gelingt, begeistert den Forscher. (7) Das Gelingen des Versuchs begeistert den Forscher. (8) *Que I 'experiment réussisse, enthousiasme le chercheur. (9) La réussite de l'expériment

enthousiasme le chercheur.

A good German —» French translation dictionary, which not only gives access to target language lexemes starting from source language items, but also describes the syntagmatic behaviour of these target language lexemes, should give this type of information. If it does not contain or give in some way access to information about the constructional properties of the French equivalent of begeistern, it does not prevent users from building up unacceptable sentences, like (8). The translation of subject clauses is not at all a special case: similar problems occur in the translation of verbs which take indirect interrogatives as complements in one language, but do not have equivalents in another language which allow for an isomorphic construction; or with verbs taking predicative complements which may or may not be realized as adjectival phrases or as nominal phrases; or with the translation of French verbs like supposer which allow for "a.c.i.'Mike constructions, whereas their German equivalents only allow for daß -clauses, etc. 6 . All these examples have one thing in common: the two languages under consideration display differences with respect to the realization of complements of verbs. However, the respective verbs are nevertheless (semantically) considered to be equivalents. We refer to this class of contrastive problems with the term "realization mismatches". Keeping track of these realization mismatches in a dictionary means • to be able to specify the full range of syntactic construction possibilities of source and target language items; • to relate individual construction types to a common "semantic" description on the basis of which it is possible to predict which constructions can replace each other, because they can be considered as "variants" (or in some sense as "synonymous"). If we try to formulate statements for the example above in (6)-(9), according to these two requirements, we have to express the following facts about begeistern and enthousiasmer. 6

A more detailed typology of such cases and examples for each of these phenomena have been

d i s c u s s e d in [HEID 1 9 8 8 B ] .

46 •

begeistern — takes a subject daß -clause; — takes a subject nominal phrase; — both realizations may be semantically related: "variants";

•

enthousiasmer — takes a subject nominal phrase;

• both verbs are semantically related; they may be considered as equivalents. We furthermore have to postulate that it is possible to consider that-clauses and nominal phrases which have nominalizations of verbs as heads as being synonymous "variants". If we represent this information in a bilingual dictionary, the most explicit way of describing the possible equivalences would be the following list 7 : • begeistern with nominal subject —» enthousiasmer

with nominal subject;

• begeistern with subject clause —> enthousiasmer with a subject nominal phrase whose head is a nominalization of the verb "of the subject clause". In cases where there exists a number of equivalents with different properties, the whole range of contextual equivalence statements possible with all of t h e m would have to be multiplied out in a contrastive dictionary entry.

2.2.2. Relations between complements of source and target language T h e following sentence from a technical text about railway security devices 8 may illustrate another contrastive problem related with the syntactic subcategorization properties of verbs: (10) Ces limitations de vitesse sont notifiées à l'avance et rappelées aux conducteurs par une signalisation latérale. 7 We give the equivalence statements in a very informal way, trying to express in ordinary sentences the facts which should be represented. 8 T h e headquarters of the Deutsche Bundesbahn, for whom we are currently evaluating the applicability of the SYSTRAN machine translation system for French — • German, provided us material from the daily translation work.

47

The entry which can be found in a one-volume French—»German dictionary s. v. rappeler (11) 9 indicates different equivalents and also gives a hint to t h e syntactic properties of source and target lexemes:

(11)

-er zurück(be)rufen, abberufen; erinnern (qc à qn jdn an etw), mahnen (qn à qc jdn an etw) ; ins Gedächtnis zurückrufen; theat herausrufen; mil einberufen; zum Sammeln trommeln od blasen; (Gesetz) aufheben; (-•• au téléphone) zurück-, anrufen; se ~~ sich erinnern (qc an etw); sich ins Gedächtnis zurückrufen ;

T h e entry in fact correlates the following two syntactic construction types: • rappeler quelque chose à quelqu 'un-, • jemanden

an etwas

erinnern.

The reader can infer from qn and jdn and from qc and an etw, respectively, that there are constructional differences.What we can not infer, however, or only through a very cooperative interpretation of those indications, is t h a t the French object (quelque chose) is translated by a no-object, whereas the German object corresponds to a dative-like d-object in French. A sentence like (10), where rappeler is passivized, can not be isomorphically translated with erinnern, since this verb does not allow for a passive form in which the equivalent of ces informations is a subject. T h e dictionary entry also gives the equivalent ins Gedächtnis zurückrufen, without however specifying the syntactic construction of this "Funktionsverbgefiige", assuming probably by a kind of implicit standard assumption t h a t , if not specified otherwise, the relation between the complements of the source and target language verbs is "parallel". T h e problem with erinnern and rappeler in (10) is that the argument realized as an object with the verb of one language, is related with a non-object with the verb of the other language. In contexts where the source language requires a passive, a m a j o r restructuring of the sentence to be translated would be necessary if one wanted to conserve the "standard" equivalent, e. g. erinnern for rappeler. a possible translation of this kind, though not at all the stylistically most fluent one 1 0 , would be: 9

This entry has been taken from [WEIS/MATTUTAT]. The restructuring implies among other thingss the introduction of a pronoun coreferenced with Geschwindigkeitsbegrenzungen, now the prepositional object of erinnern. Note also that the coordination is now at sentence level, whereas in French, two participle groups are coordinated. 10

48 (12) Diese Geschwindigkeitsbegrenzungen werden im Voraus bekanntgegeben und die Lokführer werden duch Streckensignale an sie erinnert. This phenomenon is widespread; here are some more examples from French — • German translation, where French objects are translated by non-object functions in German; the examples which we reproduce here all have only two complements: • applaudir quelqu'un - jemandem applaudieren (dative) • assister quelqu'un - jemandem helfen (dative) • aider quelqu'un - jemandem helfen (dative) • contrecarrer quelqu'un - jemandem entgegenarbeiten (dative) • cautionner quelqu'un - für jemanden bürgen (prepositional object) • cesser quelque chose - mit etwas aufhören (prepositional object) • voter quelque chose - über etwas abstimmen (prepositional object) Keeping track of these phenomena in a dictionary implies • describing explicitly the relationship between complements of source and target language verbs; • (ideally) giving "synonyms" which display different properties with respect to the relation between complements. This includes the necessity of keeping track of collocational variants, converses, etc. These phenomena, the same way as those discussed above in the context of "realization mismatches", make it necessary to have a sound approach to the description of the semantic basis on which we can decide about synonymy and "variants".

2.3. Some needs of a (machine) translation dictionary with respect to the contrastive description of syntactic properties of lexemes To sum up the requirements formulated in the discussion of the above examples, it can be postulated that • a translation dictionary should contain a detailed syntactic description of both source and target language items;

49 • it should be possible to describe partial structures with different syntactic constructions as being related, in the sense of "variants" or "synonymous constructions" ; • the contrastive description should contain explicit statements about the relations between complements of source and target language verbs. [KROMANN 1989] discusses different possibilities of presenting contrastive syntactic descriptions in a bilingual dictionary. One possibility is to multiply out all possible equivalence pairs, at the risk of redundancy. Another possibility is to set up a separate contrastive "dictionary grammar", i. e. an explicit list of contrastive types to which then individual dictionary entries make reference. To my knowledge, no bilingual dictionary makes explicitly use of the latter device, whereas the use of syntactic classes in monolingual dictionaries is widespread: dictionaries like HORNBY's Oxford Advanced Learner's Dictionary or the LDOCE, WAHRIGs monolingual German dictionaries or, in a less formalized presentation, French learner's dictionaries like DUBOIS' Dictionnaire du français langue étrangère, have sophisticated classifications of verbs according to syntactic properties. Some reasons as to why no bilingual dictionary to date has introduced a classification of translational equivalents according to types of pairings of constructions may be the following: • it does not seem a priori very easy to theoretically motivate the setup of classes which would simply relate certain construction types of source and target language. For example, the fact that certain German "datives" are translated by ¿-Objects (e. g. jmdm fehlen —• manquer à qn), is not more significant than the fact that other "datives" are rendered by "objects" (jmdm glauben —• croire qn, or jmdm danken - remercier qn); • the number of classes of construction pairs would become relatively high, since all existing combinations have to be treated and a large number of ideosyncratic pairings would have to be treated as translation classes containing only one element; • finally, the most problematic point is the definition of the basic elements of the syntactic description, since descriptive terms like "subject", "object" or "transitive" are only definable with respect to more elementary linguistic properties, like passivizability, case, pronominalization, the category of the head of a constituent appearing as a complement, and so on. These problems suggest that simply extending the use of syntactic classes as a descriptive device from monolingual dictionaries to bilingual dictionaries would

50 not be the best solution. At least the classes which would be defined would not be of major theoretic interest, and a larger contrastive dictionary grammar based on this device would become difficult to handle.

3. Approaches to the construction of bilingual dictionaries In the previous section, some problems of contrastive syntactic description for translation dictionaries have been discussed. T h e following section will be based on this collection of data; it will briefly review and compare the two most prominent approaches to the construction of bilingual dictionaries, the "directional approach" and the "concept-based approach" with respect to the requirements discussed at the end of the previous section.

3.1. T h e "directional approach" 3.1.1. Overview Since its publication in 1974, the work of LEV V . SCERBA 1 1 has also been interpreted by lexicographers who deal with germanic languages. Work by [HAUSMANN 1 9 7 7 ] ,

[KROMANN/RIIBER/ROSBACH

1984],

[KROMANN

1989],

[KROMANN/RIIBER/ROSBACH 1989] is inspired explicitly or implicitly by assumptions about the nature and use of bilingual dictionaries which by now have become more or less the common basis of discussions in this field, due to SCERBA or to discussions and interpretations of his work. These basic assumptions are mainly based on three distinctions, namely • the distinction between source language and target language of translation; • the distinction between the mother tongue of the user who is going to translate a text and the language which is less known to him, the foreign language; • the distinction between text production and text understanding or "reception". T h e consequences which have been derived from the combination of the first two pairs of opposites may be described as follows: n F o r references and a detailed discussion, see e. g. [MUGDAN 1 9 9 0 ] and [MIKKELSEN 1990], in this volume. [SCERBA 1940] has been published in a German translation in [WOLSKI 1 9 8 2 ] .

51 Given one language pair, each one of the two languages may be the mother tongue or the foreign language for a given user group; ideally, the language pair should thus be covered by four dictionaries, namely: • a dictionary A —• B for speakers of A ("Hiniibersetzungsworterbuch" according to [HAUSMANN 1 9 7 7 ] , "active" dictionary); • a dictionary A —• B for speakers of B ("Heriibersetzungsworterbuch", passive dictionary); • a dictionary B —> A for speakers of A ("Heriibersetzungsworterbuch"); • a dictionary B —» A for speakers of B ("Hiniibersetzungsworterbuch"). This typology relies on the assumption that two dictionaries of a given language pair are drastically different depending on the prior knowledge of the user with respect to source and target language. It does not so much rely on differences in the skills of text production and text reception, which is the case, however, in [MUGDAN 1990]S typology. T h e crucial point about the conclusions which have been drawn from the existence of the abovementioned four situations of translation is the following: in a situation of translation towards a less well known language (i. e. in the use of an "active" dictionary), properties of target language lexical items form constraints on the equivalence. This holds for the lexical semantic description as well as for that of any other descriptive level 12 ; this fact has probably been best described by Hans Peder KROMANN who requires that in a translation dictionary, the source language be described "with the eyes of the target language" 1 3 .

3.1.2. Syntactic information in a "directional" bilingual dictionary If the basic assumptions of the "directional" approach and the conclusions which have been drawn from them are taken seriously in the construction of a bilingual dictionary, descriptions of syntactic properties of lexemes should be "directional", too. This means that, for a given construction of a given lexeme, not only semantically adequate equivalents have to be given, but also the syntactic 1 2 W e assume that this statement is equally true for passive translation dictionaries, since we consider the translation as an inherently directional process; however the problem becomes more evident in an active dictionary, since the user, due to his lack of knowledge of the target language, cannot "cooperatively infer" knowledge which is not available from the dictionary information; in the situation of the use of a passive dictionary, this may be possible more often; thus the point is less striking with passive dictionaries. 1 3 Cf.

[KROMANN 1 9 8 9 ] ,

52

behaviour of these equivalents has to be described in detail. This includes information about isomorphic translation (whenever possible), or about constructions which have to be chosen wherever a structurally isomorphic translation is impossible. [KROMANN 1989] claims that in the active dictionary particular emphasis should be laid on differences in the syntactic behaviour of lexical items of source and target language; if an isomorphic translation is impossible, the dictionary should • issue a "warning" and explicitly state the nature of the difference; • indicate (quasi-)synonymous equivalents which allow for isomorphic translation, including support verb constructions ("Funktionsverbgefuge"), collocations, etc.

3.2. The "concept based" approach 3.2.1. Overview The authors of the proposals which I am going to classify as the "concept based" approach 14 also take as a starting point of their reasoning the distinction between source and target language, the distinction between mother tongue and foreign language and their cross-product which describes the possible translation situations. Their basic idea is that a translation dictionary for a given user group which has the mother tongue of that user group as source language, should contain as entries those facts, objects, situations, etc. which the members of the user group would like to speak about. This principle is applied in the actual dictionary making in the following way: a monolingual defining dictionary of the source language is used "as a starting point for the macro-structure" 15 of the bilingual dictionary. The monolingual defining dictionary can be seen as an inventory of pairs which are c o m p o s e d of a LEMMA a n d a n EXPLICATION 1 6 . T h i s p a i r "( LEMMA, EXPLICATION

)", which we will henceforth call a "concept", is used as a starting point for the equivalence statements contained in bilingual dictionaries with the mother tongue as a source language; if we allow ourselves a certain amount of simplification, we 14

[VAN S T E R K E N B U R G / M A R T I N / A L

1 9 8 2 ] a n d [ M A R T I N / A L 1 9 8 8 ] c o n t a i n a n o v e r v i e w of

the key ideas of the "concept based" approach. 15

16

C f . [VAN S T E R K E N B U R G / M A R T I N / A L

1982]: 228.

We do not go into details of those EXPLICATIONS here: they may be synonyms, definitions, related lexemes, etc. The actual form and contents are rather free and not bound to a formalization.

53 can assume that an equivalence statement in such a bilingual dictionary is a pair built up from a "concept" and a (set of) target language lemma(s). This architecture has been used in the construction of several bilingual dictionaries of the Dutch publisher Van Dale: • [VAN STERKENBURG/PIJNENBURG 1984] is a monolingual Dutch defining dictionary which associates with each of its entries one (or more different) semantic EXPLICATION(s), each of them illustrated with examples, etc. • There exist bilingual dictionaries with Dutch as a source language, translating into German 1 7 , French 18 and English, as well as dictionaries translating from these languages into Dutch. • The dictionaries with Dutch as a source language use the "meaning discrimination" set up by the monolingual Dutch defining dictionary. This means that all "Dutch-to-X" dictionaries contain for each Dutch lemma the same list of "concepts"; thus all three dictionaries share one and the same "inventory of meanings" composed by pairs of the form "( LEMMA, CONCEPT )". The bilingual dictionaries associate with each of these pairs one or more equivalents in the respective target language. Contextually relevant information, especially syntactic information, is linked to the equivalence statements by marks and a special system of two-digit codes.

3.2.1. Syntactic information in a "concept based" bilingual dictionary The architecture which has been described above constitutes the backbone of the first of two parts of the entries of Van Dale's bilingual dictionaries. This first part is meant as a list of "concepts" with for each "concept" a list of possible equivalent lemmas. The second part of the dictionary entry illustrates the contextually relevant properties of the target language items which appear in the equivalent lists of the first part. This information is linked through a two-digit code with the respective equivalence statements 1 9 . The second part of Van Dale dictionary entries, the contextual one, also contains illustrative material, such as examples, syntagms, etc., and translational 17

[Cox ET AL. 1986],

18

[ A L ET AL.

19

1985],

This device has been criticized by [HAUSMANN 1988], with respect to the way collocations are treated. Details on the two-digit-codes (called "cijfer-punt-cijfercode") can be found in [AL 1983] and in the introductions of the dictionaries.

54 equivalents for particular contexts. This part of the entries also encodes syntactic information about target language items. T h e way the published dictionaries make use of this device is rather problematic, however. T h e dictionary is based on a sort of implicit rule which says t h a t no extra syntactic marking of target items is given in the case of a structural isomorphism between source and target construction 2 0 . "Isomorphism" is to be taken here in t h e widest, least formal sense. A second tacit principle says t h a t no explicit marking of source language syntactic properties is needed in those cases where no contrasts with the target language have been observed by the dictionary authors, since the source language is well known to the users 21 . Besides this relative lack of information due to the fact that the authors of the Van Dale dictionaries seem to rely rather heavily on the user for this point, one could of course imagine to have a detailed description of syntactic properties of both source and target language lexemes, within the framework of the described architecture. Syntactic properties of source language items could for example be described in t h e meaning discrimination part and just be added to it 2 2 . T h e syntactic behaviour of target language lexemes could still figure in the contextual part, but eventually introduce further structure into it, since each equivalent should be described, and, ideally, all relevant syntactic properties should be enumerated. Structural differences in translation would then be described by the juxtaposition of different syntactic marks in the first part (source language lemma and meaning description) and in the second part of the entry (contextual use of target language items).

3.3. Comparing the two approaches In t h e following, we will try to compare certain features of the two approaches which have been outlined in this section. This comparison will by no means be complete. Its background are the requirements formulated in section 2.3.. These requirements concern: • the presence of detailed syntactic descriptions of both source and target 20

I would like to thank Bernard P. F. A1 for discussions about this point. He also informed me about this basic rule of presentation. 21 This is an oversimplification: very general statements like "transitive", "intransitive" are given in the first ("semantic") part of the entry. This treatment may lead to problems when different bilingual dictionaries are combined in order to produce material for the compilation of new bilingual dictionaries, as proposed by [AL 1988]. I will discuss this point in detail and make suggestions as to improvements in [HEID 1990]. 22 The meaning description might even be improved when keeping explicitly track of a more detailed syntactic classification than that currently used.

55 language lexemes; • the possibility of relating different syntactic construction types to a common semantic description ("variants"); • the possibility of stating relations between "corresponding" partial structures of source and target language, e. g. of relating complements of both languages. The first point is a problem of modularity, the second is to a large extent a semantic issue (what is the common denominator of the "variants"?), and the third point follows from the architecture of the respective dictionary, in the sense of an application.

3.3.1. Modularity The "concept based" approach is inherently modular, since the same Dutch defining dictionary serves as a starting point for the macrostructure of all bilingual dictionaries with Dutch as a source language. This modularity has been exploited by [ A l 1988] and [MARTIN 1989] who propose to reuse two different dictionaries with Dutch as a source language, say a Dutch—»French and a Dutch—»German dictionary to produce material usable for the construction of French «-»German bilingual dictionaries. The proposal is based on the assumption that, if the "concepts" translated in the individual Dutch-to-X dictionaries are constant, the translational equivalents of those "concepts" into French or German should be equivalents, too 2 3 . The results of preliminary experiments suggest that such an enterprise has reasonable chances of success. The "directional" approach, on the other hand, is not a priori conceived as being modular. Taking the constraint-oriented view seriously which is the basis of this approach, one has to assume that the categories and types of source language units to be distinguished in a bilingual dictionary are a function of the properties of the respective target language items. This means that the monolingual source and target language parts of such dictionaries are highly dependent, with respect to their structure, sense division and organization, on the language pair and the intended translation direction. The idea of having four dictionaries for each language pair is thus to be taken to mean "four different dedicated dictionaries". 23

Note t h a t this assumption is close to ideas which have been proposed in "interlingua-based" machine translation; it is radically different, however, from experiments assuming the equivalence "relation" to be transitive; see [AL 1988] for a discussion. Some of the inherent problems of this approach will be described on the basis of the results of a study conducted in collabaration with Van Dale and others in [HEID 1990].

56 The essential difference between the two approaches, as we see it, is the way user-orientation is realized; the "directional" approach most heavily insists on the difference between known and unknown information, i. e. the distinction between mother tongue and foreign language. The "concept-based" dictionary, on the contrary, insists on the idea of a constant inventory of meanings to be linguistically realized, with users' knowledge playing a less important role. In order to avoid redundancy with respect to a priori knowledge of the user, the ideal "directional" dictionary would be highly specialized and would try to leave out everything which can be deduced. On the other hand, it would insist on the most concrete, detailed and comprehensive description of phenomena which are contrastively relevant, such as "realization mismatches". Taking the knowledge of the users into account in a "concept-based" dictionary, however, may mostly come down to a question of organizing the context-specific second part of the entries and relating it with the first part.

3.3.2. Variants The question of relating syntactic constructions which a dictionary author would like to describe as being "variants" with (more or less) the same semantic interpretation is easy to answer in the "concept based" approach. If the "concepts" are taken to be a semantic representation which may be related with lexemes of different languages, there is no obstacle to the idea of relating different constructions of one language with one and the same "concept" 24 . It should be kept in mind that the existing Van Dale dictionaries are not very rich in syntactic description: but the fact remains that "variants" in the sense required in section 2.3. could be easily kept track of. Within the "directional" approach, there is also a possibility of representing "variants" : we could list all related constructions in the set of equivalents proposed for a source language item, and also integrate the variants in the source language entries. What remains implicit, however, is the pair of lemma and meaning description to which we could attach the constructions to be regarded as "variants", since no definitions, explications, etc. are explicitely given in the dictionary, or only rudimentary ones. The lexicographer who works along the lines of the "directional" approach should have an intuition about these "conceptual entities", but he is not forced to consistently write them down.

24It

would be interesting to discuss in more detail the way synonymy is treated in this ap-

proach. See for example [AL 1 9 8 7 ] .

57

3.3.3. Relating corresponding elements As it stands today, the "concept based" approach fails to provide a straightforward device for relating corresponding partial structures, such as complements of source and target language verbs. Apart from the fact that syntactic information is rather under-represented in the current Van Dale dictionaries, an integration of knowledge about relations between source and target language verb complements would not be straightforward. Currently, no explicit device is at hand, besides a rather ad hoc solution which would be to code this information in the example part. The directional approach would be more easily adaptable in such a way as to account for our problem. Since the syntactic behaviour of both languages is explicitly described, the target language specification being used as a model for the structuring of the source language, the relation between partial structures could serve as an additional structuring criterion.

4. Integrating the approaches in a computational environment: towards an architecture for a bilingual machine translation dictionary After having outlined the advantages and disadvantages of both lexicographic approaches, we now have to try to take the best out of both worlds in order to construct something new. The requirements derived from the discussion of some problems of syntactic description have been formulated above, in section 2.3. The examples used (cf. section 2.) only cover a small range of phenomena and disregard other levels of linguistic description than that of subcategorization properties of verbs. In unification based natural language processing systems, usually linguistic descriptions are represented through attribute-value descriptions (often called "feature structures"). These descriptions can be used to encode the properties of linguistic objects in a declarative way, e.g. in monolingual dictionaries. It is also possible to express conditions, for example on the cooccurrence of certain properties. This allows to make contrastive statements as they are needed within a bilingual dictionary. As more information is added to the monolingual dictionaries, the more complex the interaction of different types of information, monolingual as well as contrastive, is going to be. When we want to design a bilingual dictionary for a larger fragment, it may therefore be useful to have possibilities of structuring the different types of information which have to be represented in the monolingual and contrastive lexical descriptions. In this section, we will first give further evidence for the need for a modularization of lexical descriptions within a machine translation dictionary; the examples

58 discussed in section 4.1. will involve one more level of linguistic description than the example discussed before. The necessity of separating different types of descriptions will become evident. We will then outline the basic properties of the representational device we use and which exends the descriptive possibilities we have when using ordinary attributevalue descriptions; we make use of typed feature structures which permit the construction of hierarchies. With hierarchical representations, we can avoid redundancy in the different sets of linguistic descriptions (e.g. monolingual descriptions of the properties of linguistic objects at different levels), and it is possible to modularize the dictionaries. In section 4.3., we will then describe some of the basic elements of the lexical description which we are currently developing for French and German. This will allow to go through one of the examples discussed in section 2. and to show how we try to integrate the results from the "directional approach" and the "concept based approach" to bilingual dictionaries.

4.1. The need for modularity in a machine translation dictionary 4.1.1. The interaction between constraints from different descriptive levels The need for modularity becomes more acute when linguistic constraints from different descriptive levels come into play 25 . If we followed exclusively the "directional approach", each property or combination of properties of target language items acting as a filter on equivalence would have to be expressed through a specific contrastive statement in the translation dictionary. In [ E M E L E / H E I D / M O M M A / Z A J A C 1 9 9 0 A ] , the following situation is discussed: there is a (more or less) straightforward equivalence statement for the French German translation pair "pouvoir *-* konnen"26. On the other hand, it may be useful, within the grammatical description used in a translation system, to postulate translational equivalence between participle constructions in French and in German, irrespective of lexical items: (13) une offre enrichie