Inferring symmetry in natural language

TL;DR

This paper introduces a framework for inferring predicate symmetry in natural language, combining linguistic features and contextual models, validated on a new dataset, to enhance language model systematicity.

Contribution

It formalizes and evaluates methods for symmetry inference, proposing a hybrid transfer learning model that outperforms previous approaches.

Findings

Hybrid model best predicts empirical symmetry data

Linguistic features combined with contextual models improve inference

Symmetry inference can enhance language model systematicity

Abstract

We present a methodological framework for inferring symmetry of verb predicates in natural language. Empirical work on predicate symmetry has taken two main approaches. The feature-based approach focuses on linguistic features pertaining to symmetry. The context-based approach denies the existence of absolute symmetry but instead argues that such inference is context dependent. We develop methods that formalize these approaches and evaluate them against a novel symmetry inference sentence (SIS) dataset comprised of 400 naturalistic usages of literature-informed verbs spanning the spectrum of symmetry-asymmetry. Our results show that a hybrid transfer learning model that integrates linguistic features with contextualized language models most faithfully predicts the empirical data. Our work integrates existing approaches to symmetry in natural language and suggests how symmetry inference can improve systematicity in state-of-the-art language models.