Skip to Main Content
Social annotations are easy to adopt and widely used, but easiness leads to problems like ambiguity and noise. We present a probabilistic model that explains social tags by modeling the latent reason of each tag, namely the Tag Allocation Model (TAM), which disambiguates tags and identiﬁes the noise. We evaluate TAM ’s predictability on unseen data, and also test the model’s performance in two real world tasks: tag recommendation and tag relation discovery. More applications of TAM could be explored.