Needs-Centric Searching and Ranking Based on Customer Reviews

Online retailers have associated the introduction of user-generated product reviews with increased customer sales and decreased product returns. For all of the perceived value conveyed by customer reviews, however, little effort has been directed towards leveraging user-generated content beyond a product centric focus: customers first select a product in order to read from prior users of that product. In this paper, we integrate traditional information retrieval relevance ranking with database aggregation to model the knowledge within online product reviews and product descriptions. Customers search the knowledge base of reviews by querying on specific needs or interests. The result is a customized ranking of products, a recommendation list that is based upon and explained by the text of reviews written by prior users expressing similar needs or interests. We evaluate the approach using a knowledge base of online reviews from Epinions.com and compare results to expert recommendations from Consumer Reports.