| Summary: | 碩士 === 國立成功大學 === 資訊管理研究所 === 102 === Social network plays an important role in communication. People can discuss topics in which they are interested through social network sites (SNSs). SNSs put emphasis on instantaneous short text, so there are lots of messages every day. How to extract the hot topics from this large number of messages is a popular research issue now. Topic detection and tracking (TDT) can extract topic information from lots of mess data; however, in the past, most TDT researches focused on news corpus which is mainly long text. In many situation, the method applied for news cannot be used on SNSs since SNSs’ messages are too short and too various to extract the topics. The goal of this research is to help users quickly track the hot events most people concerned in social network. In order to get this goal, text clustering with probability and TDT are used to find the hot topics and the relation between topics. According to our experiment results, our method in the TDT task can achieve F-measure above 0.6 and it is better than the existed methods. Consequently, our research system can help users understand hot events on SNSs clearly and easily. Users can also focus on an event to track its detail information.
|
|---|
