CUCIS Technical Report Mining Online Customer Reviews for Ranking Products

The rapid increase in internet usage over the last few years has led to an extraordinary increase in electronic commerce. E-commerce web site like Amazon.com has made shopping online convenient, reliable and fast. While virtually any product can be purchased online today, it has become increasingly difficult for consumers to make their purchasing decisions based only on pictures and short description of a product. Since many online merchant sites allow customers to add reviews of the products they have bought, these reviews have become a diverse, reliable resource to aid consumers. The number of consumer reviews available has increased to an extent where it is no longer possible for a user to peruse them all manually. For example, some digital cameras sold on Amazon.com have several hundreds of reviews containing thousands of sentences. In this paper, we propose a novel text mining technique which uses customer reviews to rank products. We identify subjective and comparative sentences in reviews, and use these to build a weighted, directed product graph. This graph is then mined to find the top-ranked products. Experiments on real-world datasets show that our ranking algorithm produces results which correspond well with a manual ranking performed by