Getting BART to Ride the Idiomatic Train: Learning to Represent Idiomatic Expressions

TL;DR

This paper introduces a method to enhance BART's understanding of idiomatic expressions by integrating a lightweight adapter trained on idiomatic sentences, significantly improving its performance on idiom-related tasks.

Contribution

It presents a novel adapter-based approach to incorporate idiomaticity into BART, addressing limitations of previous models' compositional representations.

Findings

Idiomatic embeddings have higher homogeneity scores.

Sequence accuracy on idiom tasks improves by up to 25%.

Enhanced idiomatic understanding demonstrated through intrinsic and extrinsic evaluations.

Abstract

Idiomatic expressions (IEs), characterized by their non-compositionality, are an important part of natural language. They have been a classical challenge to NLP, including pre-trained language models that drive today's state-of-the-art. Prior work has identified deficiencies in their contextualized representation stemming from the underlying compositional paradigm of representation. In this work, we take a first-principles approach to build idiomaticity into BART using an adapter as a lightweight non-compositional language expert trained on idiomatic sentences. The improved capability over baselines (e.g., BART) is seen via intrinsic and extrinsic methods, where idiom embeddings score 0.19 points higher in homogeneity score for embedding clustering, and up to 25% higher sequence accuracy on the idiom processing tasks of IE sense disambiguation and span detection.