V. Claveau > Publications > TAL 42(3) Abstract OLST Version française Vincent Claveau, Pascale Sébillot, Apprentissage semi-supervisé de patrons d'extraction de couples nom-verbe, TAL (traitement automatique des langues), Hermès, Vol. 45, No. 1, 2004, Document (pdf) Abstract In the Generative Lexicon (GL) framework [PUS95], some semantic properties of nouns are expressed with the help of verbs. These noun-verb pairs are relevant in various domains, such as Information Retrieval. Their corpus-based acquisition is thus an interesting issue; moreover, discovering the contextual patterns in which these pairs occur can help to understand the GL model. This paper presents two related and fully automated techniques that allow us to acquire from a corpus both noun-verb pairs, and semantic and morpho-syntactic patterns. These techniques combine two different acquisition approaches---the statistical one and the symbolic one---and keep advantages of each approach: robustness and automation of statistical methods, quality of the results and expressiveness of symbolic ones.