Exploiting User Consuming Behavior for Effective Item Tagging
Liu Shen, Hongyan Liu · 2017
Automatic tagging techniques are important for many applications such as searching and recommendation, which has attracted many researchers' attention in recent years. Existing methods mainly rely on users' tagging behavior or items' content information for tagging, yet users' consuming behavior is ignored. In this paper, we propose to leverage such information and introduce a probabilistic model called joint-tagging LDA to improve tagging accuracy. An effective algorithm based on Zero-Order Collapsed Variational Bayes is developed. Experiments conducted on a real dataset demonstrate that joint-tagging LDA outperforms existing competing methods.