Research On Techniques Of Topic Detection ,Topic Tracking And Topic Diffusion In Forum

Forums have become a platform for people's publishing their opinions and communication. Because of the anonymity, openness and interactivity of a forum, people can issue their ideas; discuss topics they are interested in. But, on the other hand, it brings new challenges to supervision of public opinion. Topics diffuse so quickly that they can beyond our control, once a topic arouses enthusiasm in people, it diffuses in a high speed in the forums, and finally may lead to of formation of public opinion. Now, more and more researchers pay attention to BBS, and start the study on it.Feature representation of a text is important to topic detection and tracking. This paper raises a method of feature extraction based on meaningful string, and it is applied to represent a text. On the basis of extracting repeat string, we analyze the repeat strings and make them meaningful. Compared to words, the meaningful strings is semantic, independent and complete, so they can represent text effectively.Topic detection and tracking is a process of clustering and classification of posts on BBS. In this paper, we posed a method for topic detection and tracking based on meaningful string. Meaningful strings are used to represent text on the BBS, and single-pass increasing cluster algorithm is used to topic detection task, 1-NN is used to topic tracking task. We can organize topics, keep an eye on the hot topics, and follow up the subsequent reports on the topics by this process. As experiments shows, we got desirable results.We study the topic diffusion on the forums from two aspects for a better understanding of the developments of the topic in the future: firstly, do research on the topic diffusion among different forums by constructing the diffusion graph, and in this way we can express the path of diffusion directly and judge the central forums. Secondly, do research on the diffusion inside a forum and prediction. This paper explains the IDM and social-networking, which are applied to the study and prediction of topic. The experiments verify our research.