Samuel C. Gruber, Adam K. Hihalcin, Prashant Sridhar
We propose new models to induce a hierarchy over a set of tags in a social tagging system, also known as a folksonomy. We compare hierarchical clustering based solely on the tag descriptions with a method that first classifies tags into classes and then performs hierarchical clustering on the classes, and finally a method that looks at each tag as a set of posts, and considers two tags to be linked in a parent-child relationship if one tag's set of posts is close to a subset of the other's set of posts.
[ Ссылка ]
