Subjectivity grouping: learning from users' rating behavior

Hui Fang, Jie Zhang, Nadia Magnenat-Thalmann

Abstract: Considering opinions of other users (called advisors) has become increasingly important for each user in open and dynamic online environments. However, users might be subjectively different or strategically dishonest. Previous approaches on this problem generally suffer from the issue of limited (especially shared) historic experience when tracking each individual advisor's behavior. In this paper, instead, we model each advisor as part of groups by proposing a two-layered clustering approach. Specifically, in the first layer, the agent of each user clusters her advisors into different subjectivity groups and dishonest types, with respect to their rating behavior. In the second layer, each advisor is assigned to groups with respective membership degrees. Finally, each agent adopts an alignment approach to help its user align advisors' ratings to the ones of her own. Experimental results on both simulations and real data verify that our approach can better help users utilize ratings provided by advisors in opinion evaluation and recommender systems.