Using association rules for discovering tag bundles in social tagging data

Social tagging has become very popular with the rise of Web 2.0, making it an important piece of the puzzle that makes up the web phenomenon. In contrast to more elaborate ways of organizing resources, such as taxonomies or ontologies, tagging is very easy to use and understand. Because of its simplicity tagging does not create explicit, formalized, structures. By applying a well-established data mining algorithm we show how common structures in form of tag bundles (a set of related tags) can automatically be derived from a bag of tags. Based on these computed tag bundles tagging applications can assist users with improved user interfaces by introducing user based semantic labelling. For navigating and searching his or her tagged information space there are certainly better user interface elements possible than a tag cloud or a plain list of tags ordered by frequency. We propose that combining frequently associated tags into a tag bundle is such a way. We have tested our approach on a data set gained from Delicious and evaluated it by utilizing the Normalized Google Distance.