There is a growing interest on how we represent and share tagging data for the purpose of collaborative tagging systems. Conventional tags, however, are not naturally suited for collaborative processes. Being free-text keywords, they are exposed to linguistic variations like case (upper vs lower), grammatical number (singular vs. plural) as well as human typing errors. Additionally, tags depend on the personal views of the world by individual users, and are not normalized for synonymy, morphology or any other mapping. The bottom line of the problem is that tags have no semantics whatsoever. Moreover, even if a user gives some semantics to a tag while using or viewing it, this meaning is not automatically shared with computers since it’s not defined in a machine-readable way. With tagging systems increasing in popularity each day, the evolution of this technology is hindered by this problem. In this paper we discuss approaches to represent tagging activities at a semantic level. We present criteria for the comparison of existing tag ontologies and discuss their strengths and weaknesses in relation to these criteria.