Topic-sensitive Tag Ranking
出版社：
- 会议名称：The 20th International Conference on Pattern Recognition (ICPR 2010)
- 举办地点：Istanbul,Turkey
- 举办日期：August 2010
- 页数：629-632
摘要内容：
Social tagging is an increasingly popular way to de-scribe and classify documents on the web. However, thequality of the tags varies considerably since the tags areauthored freely. How to rate the tags becomes an impor-tant issue. In this paper, we propose a topic-sensitivetag ranking (TSTR) approach to rate the tags on theweb. We employ a generative probabilistic model to as-sociate each tag with a distribution of topics. Then weconstruct a tag graph according to the co-tag relation-ships and perform a topic-level random walk over thegraph to suggest a ranking score for each tag at differ-ent topics. Experimental results validate the effective-ness of the proposed tag ranking approach.
