Semantic Clustering and Similarity Calculation for Social Tagging Systems

Abstract

Social tagging systems allow Internet users to annotate resources with tags. Internet is an open system, which permits users the freedom to explore tags. However, the freedom afforded users comes at a cost: an uncontrolled vocabulary can result in tag ambiguity hindering navigation. So, a key question is how to harvest tag semantics from these systems. We present an algorithm of hierarchical tag clustering. With this algorithm, we clustering the tags into a semantic tree, then we turn every resource item into an induced tree. We propose a new method for resource retrieval, which based on semantic similarity. We present extensive experimental results on a real world dataset. When we retrieve the resource based on semantic similarity, our algorithm shows the high recall and precision. Furthermore, our algorithm demonstrates more utility for recommendation.