New Event Detection with Time Subtraction and Co-Occurring Words

T. Takaharu, T. Atsuhiro, J. Adachi, and K. Kageura (Japan)

Keywords

Text Mining, Topic Detection and Tracking, New Event Detection, Cooccurrence

Abstract

As the amount of available electronic source increase, the need for information also increases. It is almost impossible for a person to keep track all the information resource and find new events as soon as possible. We propose a new method for New Event Detection (NED). NED is concerned with developing systems that can detect the first story on a topic of interest. The method is not based on the traditional similarity measurement method, but concentrates on whether or not word co occurrence has existed. We show that the method achieves high recall for the First Story Detection (FSD) of the Topic Detection and Tracking (TDT-2) Corpus in comparison with existing methods.

Important Links:



Go Back