University of Liège

The DEFI Home Page

DEFI is an applied research project in the field of word sense discrimination and target selection in bilingual dictionaries. The project started in October 1995.

The objective is to develop a prototype system providing readers of a text in a foreign language (either French or English) with the most relevant translation of any word they select on-line, or of the multi-word unit which that word is an element of. The translations retrieved from two bilingual dictionaries are weighed and ranked from most to less relevant on the basis of the selected word’s environment in the source text (use is made, inter alia, of the collocational information provided by the dictionaries). The project also makes use of the WordNet semantic networks and of Roget’s thesaurus, as well as three monolingual dictionaries of English.

The implementation consists of two main steps:

turning the various lexical resources into Prolog databases;

devising look-up programs for both single-word and multi-word units (Prolog is used as a prototyping tool);

Lexical resources:

Oxford-Hachette and Collins-Robert En-Fr/Fr-En dictionaries;

WordNet Prolog package;

Roget’s thesaurus;

the COBUILD, LDOCE and CIDE dictionaries.


Departments involved :
English Department, French Department (I.S.L.V.), Department of Logic and Cognitive Sciences, Department of Archaeology.

Full-time researchers : M. Bourseau (computing), A. Dister (French), N. Dufour (English), E. Fraiture (logic and cognitive sciences), A. Gawrylkiewicz (computing), I. Jeanmart (archaeology), N. Leclair (English).

For an in-depth description of the English-to-French DEFI, please refer to the E-F DEFI home page

Home Page of ULg