Content Embedding Regularized Matrix Factorization for Recommender Systems

In recommender systems, numerous efforts have been made on utilizing textual information in matrix factorization to alleviate the problem of data sparsity. Recently, some of the works have explored neural networks to go for an in-depth understanding of textual item content, and further generate more accurate item latent models. These works achieve impressive effectiveness on performing recommendations. Nevertheless, there remains an open issue as how to effectively exploit description documents of both users and items in matrix factorization. In this paper, we proposed content embedding regularized matrix factorization (CERMF) to address this issue. CERMF adopts convolution neural networks to simultaneously generate the independent embedding representations for the users and the items. Then, dual embeddings are used to regularize the generation of latent models for users and items. Experimental results are proved that exploiting content information of both users and items dramatically improves the accuracy of recommendations as compared with the state-of-the-art recommendation models.