A supervised machine learning link prediction approach for tag recommendation

Abstract

One of the primary goals of tag recommendation approaches is to deal with the problem of ambiguity of tags in a folksonomy by helping users to select the most appropriate tag to annotate a resource. We propose in this work, an original approach for tag recommendation applying a link prediction using supervised machine learning. Given a user (target user) and a resource (target resource) the proposed algorithm computes a list of tags best suited for recommending target user to annotate the target resource. It first searches for users similar to the target user. Then a link prediction approach is applied on a temporal sequence of bipartite graphs coding the history of tagging of retrieved similar users. This results in obtaining one or more lists of tags for the target resource or similar resources. These lists are then merged using a list aggregation method to get a single list of tags for recommendation. The first prototype of this approach is described in this article. Preliminary results of applying the proposed approach to real dataset extracted from the bibliographical folksonomy CiteULike show the validity of the approach. © 2011 Springer-Verlag.