Research Of Topic Detection And Tracking Based On Forum

Internet Public Opinion is popular on the Internet. It is a kind of network media that shows different views on social issues and spreads statements and views through the Internet about some hot and focused problems held by public which have a strong influence and bias. With the increasing popularity and openness of the Internet, people in politics practice their rights. At the same time, there is always a part of users to spread malicious, false statements through the Internet, which could mislead the people and cause social instability. Topic detection and tracking system on the one hand can be used as a monitor on Internet forums and other forms of public opinion, and on the other hand can be used to provide users with topic classification label, which will help users to look up information they concerned.Topic detection and tracking technology based on forum topics, as it involves multiple disciplines, is still in a development stage. At present, the main approach is the cluster analysis of the forum posts to access topics. Improvement of different clustering algorithm based on different system applications is the main research directions. Existing clustering algorithms typically have certain advantages, but when applied to the system they faced with problems of dynamic data adaptation, adaptive data structures, clustering effects, clustering of data to be bound by the shape and so on. Consequently, there is no existing algorithm can fully adapt all the features mentioned above at the same time.This paper shows topic detection and tracking using text clustering which is easy to understand and easy to achieve. The core techniques of text clustering are summarized first. Then using current Vector Space Model, incremental clustering algorithm is improved from hierarchical clustering algorithm and gets the features which has more granularity. It adapt to dynamic data and update data on the incremental clustering. The algorithm is used in forum topic detection and tracking system. By testing data from two well-known forums, we can see that topic detection and tracking systems can achieve good accuracy.