Tag-Based User Interest Discovery Though Keywords Extraction in Social Network

We consider the problem of exploiting to discover user interests from social network. User tags in social networks convey abundant implications of user interests,which great benefit various tasks ranging from user profile construction to user similarity calculation based recommendation. However,user interests extraction from social tags suffer from large diversity of word choices due to different user preference,especially the words that quite specific in minority knowledge domains. In addition,the deficiency of uniform concept hierarchy and lack of explicit semantic association between tags obscure the real interests of users. To obtain user interests from tags,we propose a tag normalization algorithm based on world knowledge to underpin the construction of common tags as well as the organization of user hierarchy interest. Experiments with Sina Micro-blog (http://weibo.com/) show that our algorithm can infer user’s interests better than traditional method based on contents.