Acquiring semantics from structured corpora to enrich an existing lexicon

Lexical and semantic information is crucial for NLP by and large and for building linguistic resources in particular. However, this kind of information is hard to be obtained be it manually or automatically. The difficulties come from the very nature of the information (what do we mean by "meaning"? what is "semantics"?, how are "meanings" combined and put into words?) and from the resources (how are "meanings" formally represented and displayed?). In this paper we present a methodology for automatically acquiring semantics from structured corpora (machinereadable dictionaries and free encyclopaedias on the web). This information is used to enrich an existing lexical database to build families of words in modern French.