Tag-based recommendation systems aim to improve the search experience of the end users. However, due to different backgrounds of the end users, descriptions of the same resources may be totally different in particle size and degree of specialization, which raises the question of how to tackle the growing discrepancy of public taxonomies (Folksonomy) in the social networks. In line with this, WordNet-based similarity is used to obtain semantic distance between tags and topic categories in order to reduce the divergence of tags. This in turn improves the search accuracy. The Bayesian reasoning is introduced to infer users' preferences through mining users' comments towards particular categories. Users' interaction behavior, which may facilitate preference estimation, is considered as well to enhance search efficiency. A series of experiments are conducted based on Flickr and Delicious datasets. The results show that the proposed recommendation algorithm can effectively improve search precision and provide a greater level of user satisfaction.
|Cite as: Li, J., Li, L., Wen, X. and Liao, J. (2012). A Collaborative Filtering Recommendation System Combining Semantics and Bayesian Reasoning. In Proc. Data Mining and Analytics 2012 (AusDM 2012) Sydney, Australia. CRPIT, 134. Zhao, Y., Li, J. , Kennedy, P.J. and Christen, P. Eds., ACS. 191 - 198 |
(local if available)