Collaborative filtering with weighted opinion aspects

Collaborative filtering (CF) aims to produce recommendations based on other users' ratings to an item. Most existing CF methods rely on the overall ratings an item has received. However, these ratings alone sometimes cannot provide sufficient information to understand users' behaviors. For example, a user giving a high rating may indicate that he loves the item as a whole; however, it is still likely that he dislikes some particular aspects at the same time. In addition, users tend to place different emphases on different aspects when reaching an overall rating. This emphasis on aspects may even vary from users to items, and has a significant impact on a user's final decision. To make a better understanding of a user' behavior and generate a more accurate recommendation, we propose a framework that incorporates both user opinions and preferences on different aspects. This framework is composed of three components, namely, an opinion mining component, an aspect weighting computing component, and a rating inference component. The first component exploits opinion mining techniques to extract and summarize opinions on multiple aspects from reviews, and generates ratings on various aspects. The second component applies a tensor factorization strategy to automatically infer weights of different aspects in reaching an overall rating. The last one infers the overall rating of an item based on both aspect ratings and weights. Experiments on two real datasets prove that our model performs better compared with the baseline methods.