Semantic-Aware Techniques Enhanced Recommendations in Social Network

Abstract

This paper proposed the semantic-aware method to enhance the unknown but interesting recommendations. That is, the recommender system aims to suggest unknown but interesting groups to the target user, people should have the same interests or characteristics in the groups and the groups we recommended are high probably the target user like. The previous work, UBI, discovers unknown but interesting items based on the social behaviors of users. However, the work did not consider the information that the user really likes, e.g., the fan pages that user already joined. The information that user really likes is an important hint to recommend unknown but really interesting groups to the target user by merging social objects (such as fan pages and groups). We propose a semantic representation method of the social objects and integrate the semantic features into the UBI algorithm. Experimental results show that our method has better recommendations with unknown but user interesting items.