Creating Related Tag Groups using Co-occurrence Frequency on Blogosphere

As blogs have been successful and popular, those have generated many interesting and challenging problems to the research community. An important facet of blogosphere is that users manually annotate their contents so called tags, which describe and represent the contents of a blog page and provide additional contextual and semantic information. These tags can be useful for information retrieving and web page classification. If these tags are grouped into related set of tags, the groups can be useful for various ways. In this paper, we propose an effective method of creating semantically related tag groups using a clustering technique based on tag co-occurrence and users’ tagging tendencies in blogosphere. The related tag groups can improve the tagging experience and the utility of the tags in general. We analyze users’ tagging tendencies. Based on this, we count tags’ co-occurrence and find when tags are most related, then measure membership score between tags. This measurement is applied to clustering to create related tag groups. To verify our method, we collect tag information data from Technorati. Using this data, we perform an experiment and show our method is effective.