Discovering Event Evolution Chain in Microblog

Microblog, such as Weibo and Twitter, has become an important platform where people share their opinions. Much research has been done to detect topics and events in microblogs. Due to the dynamic nature of events, it is more crucial to monitor the evolution and trace the development of the events. People pay more attention to the whole evolution chain of the events rather than a single event. In this paper, we propose a method to automatically discover event evolution chain in microblogs based on multiple similarity measures including contents, locations and participants. We build a 5-tuple event description model specifically for events detected from microblogs and analyze their relationships. Inverted index and locality-sensitive hashing are used to improve the efficiency of the algorithm. Experiment shows that our method gain a 143.33% speed up against method without locality-sensitive hashing. In comparison with the ground truth and a baseline method, the result illustrates that it effectively covers ground truth and outperforms the baseline method especially in dealing with the long-term spanning events.