Topic Tracking Of Microblog Based On SVM And Its Application

As a key issue in the field of information processing, topic tracking has drawn muchattention. It has been applied in many fields such as public opinion analysis, digital library,etc. At present, most of the research on topic tracking system focuses on thelong text likenews, blogs, etc. The research is less on the social network information like microblog orother short text information. Recently, with the development of natural language processand machine learning technologies, the method of topic tracking system construction hasalso been developing. For the microblog information, this paper designs a method ofmicroblog topic adaptive tracking based on SVM. The biggest advantage of this method isable to automatically adaptive microblogging topic tracking, at the same time the analysisand summarization of the development of topic evolution are automatic. Only the parts ofthe original topic model training corpus collection and the building of feature words tableneed some manual process.Firstly, this paper designs a method of microblog topic adaptive tracking based onSVM. The method includes the following steps: gathering microblog information, buildingfeature words table, training classification model and analyzing the microblog topicevolvement. Building feature words table and analyzing the microblog topic evolvement,which is the main research contents in this paper.Secondly, automatic building feature words table method is studied. We adopted themethod of feature selection. The method divides into three steps: Chinese wordsegmentation, selection of feature selection index and the global weight of featurescalculation. In the step of Chinese word segmentation, adding a new word discoverymodule to improve the segmentation accuracy. Then we select the appropriate evaluationindex for screening feature words with comparing different feature selection index. Finally,we compute the global weight of feature words according to the feature selection index.Thirdly, the evolution of topic model is studied. We adopt the feedback mechanism toupdate the topic model dynamically. It is the purpose of ensure topic tracking system toeffectively track the follow-up microblog information. At the same time we use LDAmethod to extraction and summarized the new topic and check the topic whether themigration.Fourthly, we used the method of microblog topic adaptive tracking based on SVM totracking the topic of microblog automatically. And finally, we show the trajectory ofmicroblog topic evolvement. The method cantrack automatically and continuousaccurately topic microblog information and reduces the human work.