Cluster tree based hybrid semantic similarity measure for social tagging systems

As the social tagging systems becoming prevalent, it remains a critical question that how to make explicit the semantics for tags to fully facilitate Web2.0 applications. This paper establishes a cluster tree based semantic similarity measure for social tagging systems, combines it with traditional statistics based measures into a hybrid one, tailors the hybrid measure according to the effectiveness requirement of intelligent search application, and presents a case study using the empirical data retrieved from delicious website. Comparing to the traditional statistics based measures, our hybrid measure is capable of evaluating similarities between random tags even not co-occurred, can better reflect the structural influence of the network of tag co-occurrence, and is feasible for applications like intelligent search in user-centric Web2.0 environment.