A Recommendation Approach Based on Product Attribute Reviews: Improved Collaborative Filtering Considering the Sentiment Polarity

摘要

Recommender methods using reviews have become an area of active research in e-commerce systems. The use of auxiliary information in reviews as a way to effectively accommodate sparse data has been adopted in many fields, such as the product field. The existing recommendation methods using reviews typically employ aspect preference; however, the characteristics of product reviews are not considered adequate. To this end, this paper proposes a novel recommendation approach based on using product attributes to improve the efficiency of recommendation, and a hybrid collaborative filtering is presented. The product attribute model and a new recommendation ranking formula are introduced to implement recommendation using reviews. Experimental results show that the proposed method outperforms baselines in terms of sparse data.