Representing and reasoning with events from natural language

Abstract

Linguistic categories such as progressives, perfectives, tense and temporal adverbials are at the heart of our ability to describe events in natural language. Following on from the work of Moens and Steedman and the later work of Kent, we have identified a fragment of an interval tense logic of Halpern and Shoham that is expressive enough to represent the temporal readings of many simple sentences involving the linguistic categories listed above, and computable enough for entailment checking to be manageable in a reasonable time scale. We show how one can model the semantics of formulae from the fragment using simple timelines and how one can support entailment checking by comparing timelines using a simple algorithm.