关键词:  , LDA, 主题模型, 时间约束, 主题词

Abstract:  With the development of Internet, a large number of public opinion texts have been produced, and the hot topics and trends can be found by topics extraction from these texts. Because of the huge amount of the texts, and the dynamic changes of topics, a TC-LDA (Time Constrained LDA) model is proposed. TC-LDA can transform the text data into the topic vector and greatly reduce the dimension of public opinion texts, and implements the LDA’s timing conversion by adding the time constraint, which can improve the ability of LDA to capture the dynamic topic words. Experiments show that the accuracy and recall rate of TC-LDA is better than that of the similar topic model.

Key words:  latent dirichlet allocation, topic model, time constraint, topic words