Domain-Independent Detection, Extraction, and Labeling of Atomic Events

Abstract

The notion of an “event” has been widely used in the computational linguistics literature as well as in in- formation retrieval and various NLP applications, al- though with significant variance in what exactly an event is. We describe an empirical study aimed at de- veloping an operational definition of an event at the atomic (sentence or predicate) level, and use our ob- servations to create a system for detecting and prior- itizing the atomic events described in a collection of texts. We report results from testing our system on sev- eral sets of related texts, including human assessments of the system’s output and a comparison with infor- mation extraction techniques. We discuss how event detection at this level can be used for indexing, sum- marization, and question-answering.