Analysis of Topic Evolution Based on Subtopic Similarity

In the area of topic tracking, topic is developing with time. Actually speaking, traditional topic tracking approaches could track the relevant stories. However, the relation between events occurred during the topic development process could not be learned with traditional approaches. Furthermore, the whole history of the topic tracking could not be acquired either. Based on these disadvantages in traditional approaches, we propose anew method based on the calculation of the subtopic similarity.This method utilizes the idea of time partition so as to analyze the developing history of the topic. Our method concerned the time characteristic of a event as well as the location characteristic. Moreover, we use it to manipulate the information. In the experiment applying the practical data, it shows our method works more efficiently.