Can Tagging Organize Human Knowledge?

The fundamental premise of tagging systems is that regular users can successfully organize large collections for browsing and other tasks through the use of uncontrolled vocabularies. Until now, that premise has remained relatively unexamined. In this work, using library data as a guide, we look at whether tags do a good job of organizing large sets of objects. We find that tags are predominantly objective and content-based, that synonymy is not a huge problem, and that the groups of objects are well distributed in size. We find that tags often have equivalent or contained library terms. Using these equivalent and contained terms, we examine the recall of a tag on the objects it could annotate. We find that tags have high recall on popular works, but lower recall on less popular works. Lastly, we find that tags have some desirable traits for information integration, like similar top tags across systems and similar annotations of the same object across systems.