Aspect-Based Personalized Review Ranking
Huang ChunLi, Jiang WenJun · 2018
Many users of e-commerce websites are often used to making purchase decisions based on the product reviews. Since there are too many reviews attached to a product, users may not have enough time and patience to read all the reviews. Moreover, since each user has different requirements for different characteristics of the product, most reviews may have little help in making his purchase decision. To address this issue, we propose a personalized review ranking method to select useful reviews for users. To be specific, we try to assign a helpfulness score for each review, which measures the value of each review for a user's decision making. Then, we sort reviews in descending order of helpfulness, so as to generate a subset of reviews for the user. Our main idea is to find similar users' reviews for a target user, and examine the number of aspects that similar users have reviewed. By considering the similarity between users and the number of aspects of that users have reviewed, we produce personalized review recommendation according to the helpfulness of the reviews. Here, we define similar users as having similar sentimental tendencies towards the same aspects of the same products. We mainly look for similar users of similar products to improve the accuracy of review recommendation to similar users.