Improve tagging recommender system based on tags semantic similarity

Collaborative Filtering (CF), widely applied in such personalized recommender systems as e-business, e-library, is one of the most successful techniques to date. However, this recommender system based on traditional CF seems to refuse to consider user preference, resulting in the inaccuracy of recommendation. In view of the above limitations, we propose, in this paper, a new collaborative filtering method CFBTSS (Collaborative filtering base on tag semantic similarity). This approach tries to better understand user interest by analyzing the relevance between tags and items and by dealing with the problems of the similarity between words and similarity between sentences. Experiment results tested on MovieLens dataset show that CFBTSS significantly improved its recommending efficiency and accuracy compared to the traditional one, which contributes to the excellent performance of personalized recommendation system.