Co-Extracting Opinion Relations from Online Reviews Based on the Word Alignment Model

The important task of opinion mining is to mine opinion targets and opinion words from a huge number of product reviews. The one of the approach proposes a method based on partially supervised word alignment model, in which opinion relations identification is consider as an alignment process. Calculating opinion association among words is an important for constructing Co-Ranking graph; to find confidence of each candidate graph based co-ranking algorithm is used. Higher confidence candidate are extracted as opinion targets or opinion words. Prior knowledge also consider in finding confidence of candidate as being opinion target or opinion word. Previous methods are based on syntax based, compared to these methods proposed model minimizes negative effects of parsing errors. Due to use of partial supervision proposed model achieves better accuracy compared to unsupervised word alignment model. Final task is to extractive summary generation from Opinion Targets and Opinion Words with Word Alignment Model.