In this paper, the authors analyze how tags utilized by users with different motivations influence social classification effectiveness. With a very detailed description of the social library book tagging systems example, the authors point out that, in a collaborative tagging system, not all user-generated tags are created equal.

Experiments show that verbosity is the optimal feature for discriminating user behavior. It would be interesting to find out if further research can distinguish casual readers who spend five minutes on a book from serious readers who read the book from cover to cover.

This paper is one of the first to derive useful information from user bookmark tags. Using the multi-class vector space model (VSM), the proposed automated classification provides the best performance with the tags per post (TPP) measure. Librarians compare classification accuracy against results of the expert system. The automated classification, if given realistic applications, would open the Web 2.0 era to book readers, where everyone can contribute to the book shelving categorization by making their own bookmarks. This algorithm has the financial potential to save library maintenance costs, as well.