Linking Rare and Popular Tags in CQA Sites

Community Question Answer (CQA) sites are popular means for sharing knowledge in the form of questions and answers. These sites rely on tags for many purposes, such as content organi- zation, question routing, content searching, etc. Each CQA site has thousands of tags, making it challenging for the users to manually annotate their question posts with the appropriate tags. Understanding the semantic relationships between tags could aid in the tagging process, thereby properly routing the questions to the experts. Although it is relatively easier to mine the se- mantic relationships amongst the frequently used popular tags, it is difficult to do so for the less commonly used rare tags due to a lack of information about them. Most often, the rare tags are specific concepts subsumed by popular tags. For the questions to be routed to the right experts, they must be annotated with a proper mix of both popular and rare tags. In this paper, we pro- pose a novel approach to mine the semantic relationships between the rare and the popular tags. In addition, we show that the methods that are proposed to mine semantic relationships between popular tags cannot be used for rare tags. Specifically, we identify the top-k popular tags that are semantically related to a given rare tag, which is done using a set of semantic and topological features. Extensive evaluations on CQA datasets show the superiority of our proposed method over state-of-the-art methods.