Online voting is an emerging feature in social networks, in which users canexpress their attitudes toward various issues and show their unique interest.Online voting imposes new challenges on recommendation, because the propagationof votings heavily depends on the structure of social networks as well as thecontent of votings. In this paper, we investigate how to utilize these twofactors in a comprehensive manner when doing voting recommendation. First, dueto the fact that existing text mining methods such as topic model and semanticmodel cannot well process the content of votings that is typically short andambiguous, we propose a novel Topic-Enhanced Word Embedding (TEWE) method tolearn word and document representation by jointly considering their topics andsemantics. Then we propose our Joint Topic-Semantic-aware social MatrixFactorization (JTS-MF) model for voting recommendation. JTS-MF model calculatessimilarity among users and votings by combining their TEWE representation andstructural information of social networks, and preserves thistopic-semantic-social similarity during matrix factorization. To evaluate theperformance of TEWE representation and JTS-MF model, we conduct extensiveexperiments on real online voting dataset. The results prove the efficacy ofour approach against several state-of-the-art baselines.
展开▼
