LDA-based user interests discovery in collaborative tagging system

Abstract

The success and popularity of collaborative tagging systems, such as delicious1, Flickr2, Last.fm3, has increasingly centered on. Users of these websites can easily tag their interested WebPages, photos and music with their preferred words. Subsequently, the extensive tagging data attract many researchers to mine useful information from these. In this paper, we propose a novel user interests quantified approach based on user-generated tags. Moreover, by means of the generative probabilistic model Latent Dirichlet Allocation (LDA), we acquire the interests for each user. Experimenting with the dataset provided within the ECML PKDD Discovery Challenge 2009, our method makes better performance.