A Survey of Statistical Topic Model for Multi-Label Classification

Much of texts embedded in Web is annotated with human interpretable labels, such as tags on web pages and subject. Statistic topic model for multi-label classification is a power technology to handle the multi-labeled textual data at the word level. However, standard topic model is a completely unsupervised algorithm. Therefore, the key of incorporating supervised label set into its topic modeling procedure is to establish the relationship between topics and labels. In this paper, multi-label topic model is summarized by analysis of existing studies; especially, on the basis of relationship between topics and labels, we describe four categories of multi-label topic model, and their reprehensive models. To the best of our knowledge, this is the first effort to review the development of multi-label topic models.