CFUI : Collaborative Filtering with Unlabeled Items

As opposed to Web search, social tagging can be considered an alternative technique tapping into the wisdom of the crowd for organizing and discovering information on the Web. Effective tag-based recommendation of information items is critical to the success of this social information discovery mechanism. Over the past few years, there have been a growing number of studies aiming at improving the item recommendation quality of collaborative filtering (CF) methods by leveraging tagging information. However, a critical problem that often severely undermines the performance of tag-based CF methods, i.e., sparsity of user-item and user-tag interactions, is still yet to be adequately addressed. In this paper, we propose a novel learning framework, which deals with this data sparsity problem by making effective use of unlabeled items and propagating users’ preference information between the item space and the tag space. Empirical evaluation using real-world tagging data demonstrates the utility of the proposed framework