A New Method of Topic Tracking for Micro-Blog Texts Based on Semantic Relevance

For the sparseness of the micro-blog text and the less number of micro-blog texts related to the original topic model, a new method of topic tracking for micro-blog texts based on semantic relevance is proposed. Firstly, it constructs a semantic relevance model by using HowNet, mutual information and such structured information as forwarding, comments among micro-blog texts. Secondly, the mode is used to measure the relevance between micro-blog texts and topics, which are expressed in the form of a list of key words, thus comprehensively measuring the relevance between them from the text content and structured information. Finally, it is applied into micro-blog text data set to realize topic tracking. Experimental results show that compared with other micro-blog topic tracking methods, this method makes the miss rate and false detection rate further decreased, effectively improving the effect of topic tracking.