Automatic Summarization of Opinionated Texts Résumé automatique de textes d'opinion

In this paper, we present a summarization system that is specifically designed to process blog posts, where factual information is mixed with opinions on the discussed facts. Our approach combines redundancy analysis with new information tracking and is enriched by a module that computes the polarity of textual fragments in order to summarize blog posts more efficiently. The system is evaluated against English data, especially through the participation in TAC (Text Analysis Conference), an international evaluation framework for automatic summarization, in which our system obtained interesting results.