Disentangling Multi-Facet Social Relations for Recommendation
Xiao Sha, Zhu Sun, Jie Zhang
Abstract
Social networks have proven to be effective for high-quality item recommendation. Most social recommenders, however, are insufficient to capture the user preferences over items, as they largely neglect the multi-facet social relations latent in the social network, such as classmates, colleagues, and families. We, therefore, propose a novel disentangled social recommendation (DSR) framework to exploit the multifacet social relations for enhanced item recommendation. Specifically, DSR explicitly disentangles the social relations into multiple facets, and encodes the social influence under each facet into disentangled user embeddings in the social network. The multiple user embeddings are then aggregated via a facet-level attention mechanism, which distinguishes the effective facets for better inferring user interests over items. Extensive experiments show the superiority of DSR against the state-of-the-art methods and its potential in alleviating the data sparsity issue.