Sentiment classification of review documents using phrase patterns

Sentiment Analysis determines the polarity of text whether it belongs to positive or negative polarity. One motivation for sentiment analysis research is the need for user and e-commerce companies to know the public opinion from blogs, online forums, reviews about certain products, services, topics etc. Phrases are important in extracting contextual information which is important for sentiment classification. Phrases can convey sentiment information more efficiently than individual words. In this paper, sentiment-rich phrases are extracted using Part-of-speech (POS) based rules and dependency relation in the document that are capable of extracting contextual and syntactic information from the document. Next, semantic orientations of these phrases are calculated using Point-wise Mutual Information (PMI) based method. Finally, review document is classified after aggregating the semantic orientation of all the phrases into positive or negative polar document.