Hypoth` eses pour la construction et l'exploitation conjointe d'une base lexicale s´ emantique bas´ ee sur les vecteurs conceptuels

In the framework of meaning representation in NLP, the TAL research group at LIRMM develops a system for analysing thematic aspects of texts and for lexical disambiguation based on conceptual vectors. These vectors aim at representing sets of ideas related to any kind of textual segment (words, phrases, sentences, texts). Associated to various information, lexical functions for instance, vectors allow to represent the semantic of such textual segments. In this paper, we expose the hypothesis grounding the construction of a semantic lexical database. We present the overall organisation of such a system and an example of cooperation between agents in the context of a lexical item learning.