A Novel Spatial Tag Cloud Using Multi-Level Clustering

Abstract

Conventional tag cloud systems can only present the frequency of tags and the connection of tags and tag clusters, but cannot properly express the strength of the relationships. In this paper we tackle this problem by improving the representation of tag clustering. We combine the advantages of conventional systems to create a multi-level interactive tag cloud system using clustering scheme. In our tag cloud system, tags are mapped onto a two-dimensional space. A force-directed algorithm is applied to make the tags move in 2D space. The tags will move until equilibrium is reached. As a result, the tags with high correlation are clustered in the close vicinity, and so are the tag clusters with high correlation. In addition, we let the users to interact with the tag cloud by zooming and panning for different perspectives and let user to select a tag than the tag will locate at the center of window and show the correlations of the user selected tag with other tags. We believe that our system can improve the user browsing experience, and can also be served as a great basis for social network analysis and other related research.