Group Division for Recommendation in Tag-based Systems

The common usage of tags in these systems is to add the tagging attribute as an additional feature to re-model users or resources over the tag vector space, and in turn, making tag-based recommendation or personalized recommendation. With the help of tagging data, user annotation preference and document topical tendency are substantially coded into the profiles of users or documents. However, obtaining the proper relationship among user, resource and tag is still a challenge in social annotation based recommendation researches. In this paper, we utilize the relationship from between tags and resources and between tags and users to extract group information. With the help of such relationship, we can obtain the Topic-Groups based on the bipartite relationship between tags and resources; and Interest-Groups based on the bipartite relationship between tags and users. The preliminary experiments have been conducted on MovieLens dataset to compare our proposed approach with the traditional collaborative filtering recommendation approach approach in terms of precision, and the result demonstrates that our approach could considerably improve the performance of recommendations.