UniNE at TREC 2008: Fact and Opinion Retrieval in the Blogsphere

摘要: Abstract : This paper describes our participation in the Blog track at TREC 2008 evaluation campaign. The goes beyond simple document retrieval, its main goal is to identify opinionated blog posts and assign a polarity measure (positive, negative or mixed) these information items. Available topics cover various target entities, such as people, location product for example. year's task may be subdivided into three parts: First, retrieve relevant (facts & documents), second extract only documents (either positive, third classify having positive polarity. For first part of we evaluate different indexing strategies well retrieval models Okapi (BM25) two derived from Divergence Randomness (DFR) paradigm. opinion detection part, use approaches, an additive logistic-based model using characteristic terms discriminate between classes.