Abstract

The popularity of social media creates a large amount of user-generated content, playing an important role in addressing cold-start problems in recommendation. Although much effort has been devoted to incorporating this information into recommendation, past work mainly targets explicit feedback. There is still no general framework tailored to implicit feedback, such as views, listens, or visits. To this end, we propose a sparse Bayesian content-aware collaborative filtering framework especially for implicit feedback, and develop a scalable optimization algorithm to jointly learn latent factors and hyperparameters. Due to the adaptive update of hyperparameters, automatic feature selection is naturally embedded in this framework. Convincing experimental results on three different implicit feedback datasets indicate the superiority of the proposed algorithm to state-of-the-art content-aware recommendation methods.