A Recommender System based on Detected Users' Complaints by Analyzing Reviews
Toshinori Hayashi, Yuanyuan Wang, Yukiko Kawai, Kazutoshi Sumiya · 2018
Even though the popularity of e-commerce recommender systems continues to spread, traditional recommender systems might not recommend alternatives that can solve the problems of the items flagged by users. Since users need to choose options on existing services, e.g., screen quality and size, it remains difficult to satisfy their requirements. Therefore, we propose a novel item recommender system based on our analysis of complaint data and review comments on e-commerce. Our system first generates negative feature vectors from complaint data and positive feature vectors from review comments. Next, it calculates the similarities of these two vectors and identifies reviews that can solve the complaints about the items checked by users, and provides alternatives for each user complaint. In this paper, we describe our proposed recommendation method based on complaint and review data.