Unsupervised Discovery of Unaccusative and Unergative Verbs

TL;DR

This paper introduces an unsupervised approach to identify unaccusative and unergative verbs in English by generating sentence variants and probing language models, avoiding the need for annotated data.

Contribution

The method is novel in using unsupervised generation and probing techniques to classify verbs without relying on semantic role annotations.

Findings

Achieved results comparable to supervised approaches

Does not depend on annotated resources

Effective in identifying causative-inchoative verb classes

Abstract

We present an unsupervised method to detect English unergative and unaccusative verbs. These categories allow us to identify verbs participating in the causative-inchoative alternation without knowing the semantic roles of the verb. The method is based on the generation of intransitive sentence variants of candidate verbs and probing a language model. We obtained results on par with similar approaches, with the added benefit of not relying on annotated resources.