Mining Online Customer Reviews for Product Feature-Based Ranking

Abstract

Recent trends have indicated that large numbers of customers are switching to online shopping. Online customer reviews are an unbiased indicator of the quality of a product. However, it is difficult for users to read all reviews and perform a fair comparison. We describe a methodology and algorithm to rank products based on their features using customer reviews. First, we manually define a set of product features that are of interest to the customers. We then identify subjective and comparative sentences in reviews using text mining techniques. Using these, we construct a feature-specific product graph that reflects the relative quality of products. By mining this graph using a page-rank like algorithm (pRank), we are able to rank products. We implement our ranking methodology on two popular product categories (Digital Camera and Television) using customer reviews from Amazon.com. We believe our ranking methodology is useful for customers who are interested in specific product features, since it summarizes the opinions and experiences of thousands of customers.

Keywords: Text mining, page-rank, quality.