Analysis of online reviews for evaluating informative reviews

This paper proposes an evaluation method for informative reviews depending on personal values, and shows results of analyzing actual online reviews. Reviews of items are posted by users on many shopping sites. A user who is considering purchasing an item will refer to reviews about it. However, it becomes overloads if there are so many reviews for the item. In particular, there exist reviewers who post reviews of the same contents to multiple items, which force a burden on readers. Proposed method first removes copy&paste reviews, from remaining reviews expressions mentioning evaluative attributes are extracted. Informative reviews are determined based on several criteria, which include reader's personal values. This paper reports the result of analyzing actual online reviews, based on which feasibility and challenges of the proposed method are discussed.