Anticipatory event detection via classification

The idea of event detection is to identify interesting patterns from a constant stream of incoming news documents. Previous research in event detection has largely focused on identifying the first event or tracking subsequent events belonging to a set of pre-assigned topics such as earthquakes, airline disasters, etc. In this paper, we describe a new problem, called anticipatory event detection (AED), which aims to detect if a user-specified event has transpired. AED can be viewed as a personalized combination of event tracking and new event detection. We propose using sentence-level and document-level classification approaches to solve the AED problem for some restricted domains; given some user preferred topic event transition, we first train the corresponding event transition model, and then detect the occurrence of the transition for the stream of news covering the topic. Our experimental results on both terrorist-related and commercial events demonstrate the feasibility of our proposed AED solutions.