Aspectuality Across Genre: A Distributional Semantics Approach

TL;DR

This paper demonstrates that distributional semantics can effectively model lexical aspectual classes of verbs, outperforming previous methods and revealing genre and discourse influences on telicity.

Contribution

It introduces a novel distributional semantics approach to model aspectual classes and provides a new annotated dataset linking telicity with genre and discourse.

Findings

Local context is most indicative of aspectual class

Closed class words are stronger discriminators than content words

Approach outperforms previous methods on three datasets

Abstract

The interpretation of the lexical aspect of verbs in English plays a crucial role for recognizing textual entailment and learning discourse-level inferences. We show that two elementary dimensions of aspectual class, states vs. events, and telic vs. atelic events, can be modelled effectively with distributional semantics. We find that a verb's local context is most indicative of its aspectual class, and demonstrate that closed class words tend to be stronger discriminating contexts than content words. Our approach outperforms previous work on three datasets. Lastly, we contribute a dataset of human--human conversations annotated with lexical aspect and present experiments that show the correlation of telicity with genre and discourse goals.