关键词: 非负矩阵分解，标签，标签语义挖掘

Abstract: With the development of Web2. 0 technologies, social tagging systems are becoming more and more popular,which makes tags widely used to retrieve, categorize, and manage users' collections. However, people arc free and uncontrollable to use tags, resulting in a large number of tags that are redundant, unclear in semantics. To deal with this problem, we proposed a tag semantics mining algorithm based on non-negative matrix factorization method. We got a tag subspace containing a series of semantic related tag-bases by factorizing tagged data of users using non-negativity constraints,to make synonymous and related tags into the same tag-basis,and categorize polysemous tags into different semantic tag-bases. Simultaneously, the tasks of grouping synonymous tags and distinguishing polysemous tags were done by the proposed approach. A large number of experiments demonstrate the effectiveness of the proposed algorithm on mining tog semantics.

Key words: Non-negative matrix factorization, Tag, Tag semantics mining