2011/003 - Un système de détection d’opinions fondé sur l’analyse syntaxique profonde
TALN, Montpellier, France, June 27th-July 1st, 2011.
In this paper, we present an opinion detection system built on top of a deep robust syntactic parser. The goal of this system is to extract opinions associated to products but also to characteristics of these products, i.e. to perform feature-based opinion extraction. To carry out this task, and following the results of a target corpus study, the robust syntactic analyzer is enriched by the association of polarity to pertinent lexical elements and by the development of generic rules extracting semantic relations of opinions, in order to feed an opinion representation model. A first evaluation gave very encouraging results, but many perspectives remain to be explored.