Mining latent tag group based on tag dependency relation for recommendation in collaborative tagging systems

Currently, collaborative tagging systems have been applied in recommendation systems [2] on a large scale, during which the analysis of tags group is unavoidable. However, a sparsity problem is interfering with most of the current collaborative tagging systems, and there are only a few folks using a considerable number of tags to describe one resource. Through the deep investigation to the statistical relation between the tags in collaborative tagging systems, a mining method named BTDR based on the latent dependence between tags is proposed in this paper. We also conduct experiments to evaluate the proposed method.