Collective Matrix Factorization using Tag Embedding for Effective Recommender System

Many people communicate each other through online community, SNS as Instagram, Facebook, etc. Most of these services annotate on their clips or pictures by using tags, which contain some information and can describe their contents. In this paper, we propose a new recommender system using word embedding with tag information and collective matrix factorization technique. By vectorizing tags that users annotated, we make user-tag matrix by merging tag vectors and factorize it together with user-item matrix. We show that this method effectively works through experiments.