del.icio.us is a website for “social bookmarking” where users can store and access their bookmarks online, along with descriptive keywords or “tags.” When a user of del.icio.us logs in to their account and adds a bookmark, she may also tag that bookmark with any 10 or fewer single words that she feels are somehow related to that web page. Both the tags and the bookmarks are then publicly available; searching by a tag produces all of the bookmarked web pages ever tagged with that word. Because the tags are public, it is possible that users’ choices regarding what tags to apply could be influenced by the tagging practices of others, and a consensus might emerge for which tags should be used in a given context. However, it has long been accepted that people use language imprecisely, and meaning is negotiated on-the-fly during conversation. This imprecision is evident not only in communication, but also when people are asked to create keywords for recipes and names for common editing operations, and when user-generated index terms are compared with Library of Congress subject headings. In fact, the probability that two people will generate the same label for the same object is widely held to be less than 20%.

A question remains about whether users of del.icio.us practice social or selfish tagging. An analysis of bookmark, user and tag data for 349 web pages downloaded via del.icio.us was conducted to discover whether the “vocabulary problem” is present in the way users select tags for web pages. Results indicate that there is very little inter-user agreement, suggesting that most users consciously or inadvertently tag selfishly. These tagging practices have important implications for the findability of web pages in del.icio.us.

Download: PDF

---