Term extraction from web reviews with opinion heuristics

Term extraction from Web reviews should obviously be benefitted from the large volume of reviews on products or services. This paper addresses the opinion heuristics including opinion patterns and opinion collocations. The opinion target network is proposed in this paper to model the opinion patterns in a two-layer direct graph. Meanwhile, the unified collocation framework is presented to model the opinion collocations. With the opinion target network, a term extraction method is designed to locate opinion terms in a bootstrapping process. A term filtering method is also implemented based on the unified collocation framework. Experimental results show that our method for term extraction is promising.