Coping with vague and fuzzy words : a Multi-Expert Natural Language System which overcomes ambiguities

the system we present in this contribution is a sub-part of an extensive natural language processing architecture. It is devoted to a particular phenomenon in natural language : the polysemy of words, which produces effects such as vagueness and fuzziness and impedes the normal automatized processing of sentences. We try to show in this paper how our architecture, currently running on a network of SUN stations, relates to preceding works wellknown in NLP, and what particular improvement this system for polysemy resolution provides. This system benefits from the dedicated properties of the theoretical model for lexical description. The latter derives its power from the coupling of non monotonic logic (which proved to be very useful for semantic interpretation) and a multi-valued approach for vagueness handling.