Social tagging constitutes one of the defining characteristics of Web 2.0 as it allows users to collectively classify and find diverse resources, such as Web pages, songs or pictures, using open-ended tags. The data structures underlying these systems, also known as folksonomies, suffered an explosive growth on account of the widespread success of social tagging. Thus, it is becoming increasingly difficult for users to find interesting resources as well as filter information streams coming from this massive amount of user-generated content on Web 2.0. In addition, most resources lacks easily extractable content to apply traditional content-based profiling approaches. In this paper we present an approach to build tag-based profiles for multimedia resources (such as songs, pictures or videos) using the social tags associated to resources as a means to describe them and, in turn, user interests. Experimental results show that the tags assigned by members of the community can help to predict the interestigness of a given resource for a user in an effective way.