Combinaison de modèles de langage pour l'identification de thèmes

Brigitte Bigi (1) , Renato de Mori (1) , Marc El Bèze (1) , Thierry Spriet (1)

1 LIA - Laboratoire Informatique d'Avignon

A new statistical method for Language Modeling and spoken document classification is proposed. It is based on a mixture of topic dependent probabilities. Each topic dependent probability is in turn a mixture of n-gram probabilities and the probability of Kullback-Lieber (KL) distances between key-word unigrams and distribution obtained from the content of a cache memory. Experimental result on topic classification using a corpus of 60 Mwords from the French newspaper Le Monde show the excellent performance of the cache memory and its complementary role in providing different statistics for the decision process.

Mots clés: Topic identification

Domaines: Informatique et langage [cs.CL]; Sciences de l'information et de la communication

Brigitte Bigi, Renato de Mori, Marc El Bèze, Thierry Spriet. Combinaison de modèles de langage pour l'identification de thèmes . XXIIèmes Journées d'Etudes sur la Parole, 1998, Martigny, Suisse. pp.347-350. ⟨hal-01392234⟩