Temporally Anchoring and Ordering Events in News
Inderjeet Mani, Barry Schiffman · 2004
This paper describes a domain-independent approach to temporally anchoring and ordering events in news. The focus is on event-event and event-time linking. The approach involves mixed-initiative corpus annotation, with automatic preprocessing to identify clause structure, tense, aspect, and temporal adverbials. A controlled experiment reveals the capabilities of humans in ordering events in news. The paper then develops and evaluates two different approaches to machine learning of ordering information: learning from event-event ordering judgments, and learning from event-time anchoring judgments.