STMF: A Sentiment Topic Matrix Factorization Model for Recommendation

Traditional recommender system model latent matrix factorization only use the ratings but ignore the information hidden in reviews text. In recent years, there have been some models based on latent matrix factorization exploiting reviews. Most of them use the topics in reviews because the topics in reviews can capture item features well. However, they missed the sentiment contained in reviews while the sentiment hidden in reviews reflects user preference. In this paper, we propose a novel matrix factorization model which simultaneously considers sentiment and topics involved in reviews and ratings as well. Experimental results on real datasets show that our model reached the performance of state of the art models, and our model has better interpretability especially in user preference.