A Method of SNS Topic Models Extraction Based on Self-Adaptively LDA Modeling

While SNS(Social Network Services) playing an increasingly important role in today´s online world, SNA(Social Network Analysis) and text mining based on such communication are becoming more and more useful for a wide variety of applications. However, topic models, which have been widely used in information classification and retrieval, are not proper for some SNS such as microblogging. Moreover, It is also quite important but difficult to select topic number for a specific target. In this paper, first we present a new evaluation metric for topic models extraction from SNS dataset. Then, by combining LDA modeling with this metric, a self-adaptively LDA modeling method is proposed. Experiments were successfully performed, and the results show that the proposed LDA modeling method can self-adaptively reach the appropriate number of social topics without losing performance for microblogging-like dataset.