Automobile, Car and BMW: Horizontal and Hierarchical Approach in Social Tagging Systems

摘要

Social tagging systems have recently emerged as an effective way for users to annotate and organize large collections of resources on the Web. Moreover, they also facilitate an efficient sharing of vast amounts of resources among different users. In this paper, we analyze tags' usage pattern in real world data sets and find that among tags representing the same concept, some tags are less popular, resulting in reduced exploring effectiveness in the current social tagging systems. Another limitation is that users cannot roll-up or drill-down the concept hierarchy of tag queries, resulting in the limited scope of service and a failure to meet users' dynamic information needs which often change with the current information provided. In order to overcome these shortcomings, we propose a novel three-phase approach as the following: (1) finding semantically-related tags for tag query; (2) constructing clusters of tags representing the same concept; and (3) building hierarchical relationships among clusters of tags. Based on our approaches, horizontal and hierarchical exploration can be implemented. Experiments employing real world dataset show encouraging results and confirm that the proposed approaches are very effective.