Finding Semantic Orientation of Reviews Using Unsupervised PMI Algorithm

Recent years have shown quick expansion of the social web over the Internet, where individuals can express their opinion on various things, for example, products, persons, subjects, and discussion etc. As e-commerce is quickly developing, item audits on the Web have turned into a critical data hotspot for clients' choice making when they want to purchase items on the web. Sentiment classification of such reviews of individuals generally requires lot amount of training data but availability of labeled data for different domains is generally time consuming and tedious task. This paper present simple unsupervised learning algorithm called Pointwise Mutual Information (PMI) followed by Semantic Orientation (SO). Averaging the semantic orientation of phrase does the classification of user reviews. Phrase with positive semantic orientation is associated with positive sentiment and negative semantic orientation is associated with negative sentiment. If the average semantic orientation of phrases is positive then the review is classified as Positive otherwise Negative.