ABSTRACT
The Internet has become the universal information source, allowing people from all over the world to share their personal experiences through participation in online forums, wiki pages, blogs, etc. This sharing invariably includes user opinions about products, services, persons, and organizations, among many other things. The sharing of personal experiences regarding a product or service has become the epitome of user empowerment. Internet sites such as Amazon.com (pioneer in online book reviews) and eBay.com (first to popularize buyer/seller ratings) have propelled into the leagues of major online shopping brands simply because they understand how important user reviews/opinions affect people’s buying decisions. Smaller sites specializing in a particular product genre have also blossomed. For example, instead of walking into a retail store to make an impulsive digital camera purchase, people have learned to educate themselves by visiting camera review sites such as dpreview.com, steves-digicams.com, and others, before making the plunge to actually buy a digital camera. Naturally, first-hand evaluations or comments from customers who have used a product are extremely valuable. More so a large number of customer opinions/reviews can be assembled and summarized in a neat fashion. However, it is not easy to compare similar products by manually reading online posts/comments that could add up to the hundreds for some popular brand or model. Unfortunately, automatic approaches for analyzing consumer opinions suffer from the succinctness of typical online comments, which cannot be accurately represented using classical vector space models (VSM).9 We thus propose a novel approach to represent individual sentence segments using a word adjacency matrix, which will greatly facilitate the automatic sentiment analysis of hundreds and thousands of user opinions, ultimately simplifying the task of data collection for market analysts and consumers alike.
