Feature Extraction from Product Reviews using Feature Similarity and Polarity

Research on developing techniques to access user generated content, and specifically user reviews on different products, came in the focus of the information research community in recent past. In particular, this paper addresses the problem of extracting the features from user comments of a particular product, taking advantage of a corpus with a semi-structured format: pros, cons and summary. In this paper we propose a technique to extract a set of features based on user generated pros and cons for a particular product. Then using this set we test a feature similarity function to obtain new features from reviews (both from the pros/cons and from the free text summary) of the same and other products. Our experimental results have shown interesting conclusions.