Integrating Reviews into Personalized Ranking for Cold Start Recommendation

Item recommendation task predicts a personalized ranking over a set of items for individual user. One paradigm is the rating-based methods that concentrate on explicit feedbacks and hence face the difficulties in collecting them. Meanwhile, the ranking-based methods are presented with rated items and then rank the rated above the unrated. This paradigm uses widely available implicit feedback but it usually ignores some important information: item reviews. Item reviews not only justify the preferences of users, but also help alleviate the coldstart problem that fails the collaborative filtering. In this paper, we propose two novel and simple models to integrate item reviews into matrix factorization based Bayesian personalized ranking (BPR-MF). In each model, we make use of text features extracted from item reviews via word embeddings. On top of text features we uncover the review dimensions that explain the variation in users’ feedback and these review factors represent a prior preference of a user. Experiments on real-world data sets show the benefits of leveraging item reviews on ranking prediction. We also conduct analyses to understand the proposed models.