New Event Detection based on Temporal Topic Model
Qiaoming Zhu · Intelligent Computer and Applications · 2011
New event detection(NED) is aimed at detecting the one being reported on a new event from one or multiple streams of news stories.Preliminary experiments show that the particular time of the events' occurrence is one of the important attributes that form the event.This paper pro-poses the Temporal Topic Model(TTM).TTM divides the topics and news into some events according to different time expressions which describe the particular time of the events' occurrence.TTM estimates probability of the time expressions as the key event or dependent novel event based on the fr-equency and growth rate in topics.The time expressions' frequency and growth rate are also used for measuring the weight of event in the relevance matching.On that basis,NED relies on time behavior to suppress the events matching occur in different time quickly and adjusts the weight distribu-tion of relevance determination with the key event and novel event.