Classification Models for New Event Detection
G. Mayil Muthu Kumaran, James W Allan, Andrew McCallum · ScholarWorks@UMassAmherst (University of Massachusetts Amherst) · 2004
New event detection (NED) involves monitoring news streams to detect the stories that report on new events. In this paper we explore the application of machine learning classification techniques for this task. We introduce the concept of triangulation with illustrative examples. We develop new features that build on this concept, and the named entities present in a document. The classifiers we developed showed significant and consistent improvement over the baseline vector space model system, on all the collections we tested on. Analysis of the performance of our classifiers suggests the utility of named entities, and the applicability of machine learning techniques to the NED task.