An Interlingual Lexical Organisation Based on Acceptions From the PARAX mock-up to the NADIA system

Many projects are conducted to develop multilingual lexical databases. Some of these projects use an interlingual approach (KBMT-89, EDR, ...), where others choose a bilingual approach (Multilex, ...). This paper presents an interlingual approach based on acceptions (word-senses) aiming at the development of a multilingual lexical database management system: NADIA. With this approach, the interlingual lexicon is the union of the acceptions of the languages that appear in the database. The interlingual set of relations is freely defined by the linguist, with the exception of the predefined “is-a” relation, necessary because acceptions of terms from one language don’t always have lexical equivalents in another language. To illustrate this lexical organisation, we show in detail a mock-up build as a preliminary experimentation: the PARAX mock-up which handles 5 languages: French, English, German, Russian and Chinese. This lexical organisation has no influence neither on the linguistic content of the monolingual dictionaries, nor on the format adopted by the linguist to encode the information. The NADIA system is designed to be multilingual (multilingual databases can be handled), application independent (databases can be used for several purposes), generic (the linguistic structures are defined by the linguist) and theory independent (many computational formalisms can be used to define the linguistic structures). The system provides many tools (browser, editor, defaulter, coherence checker, ...) to simplify the management of a multilingual lexical database. Moreover, it handles the management of the interlingual dictionary as much as possible. Finally, we justify our choices by a detailed study of existing projects on multilingual lexical databases and some of the problems they have had to deal with: EDR and KBMT-89 projects (which use a knowledge based approach) are faced with theoretical and methodological problems (concept discrimination, knowledge representation, concept classification, ...). The acception based approach is a good choice to avoid the complexity introduced by knowledge representation problems and to keep the advantages of the interlingual approach. Multilex provides a way for the linguist to define the different linguistic structures, however, the linguist can only use typed feature structures to code his linguistic theory. We have chosen to provide the linguist with different formalisms (trees, graphs, typed feature structures, sets, ...).