Conceptualization of a lexical information system, a skeletal interface for NLP
Résumé
Lexical ressources which would be normalized and freely accessible is a major issue in the NLP research area. This article aims to show how to built an information system which allow (1) a freely access for distinct NLP systems (tagging, parsing, information extraction, etc.) and (2) an easy update of data by a restricted team of researchers, this update being manual or computed. Starting with a subset of the MULTEXT lexicon which is accessible from the server ABU, we aim to take into account the various needs and the variability of accessible lexical data. Our modelisation is evaluated with three existing systems of our team : EGAL (parsing), a builder of Tag grammars and VULCAIN (information extraction).