IEKG: A Commonsense Knowledge Graph for Idiomatic Expressions

TL;DR

This paper introduces IEKG, a knowledge graph for idiomatic expressions that enhances language models' understanding of figurative language by encoding commonsense knowledge, leading to improved comprehension and generalization.

Contribution

The work constructs IEKG, extending ATOMIC2020, and demonstrates how pre-trained language models can be converted into knowledge models for better idiomatic expression understanding.

Findings

PTLMs can be converted into knowledge models with IEKG

IEKG quality verified through automatic and human evaluation

Enhanced IE comprehension and generalization in NLP tasks

Abstract

Idiomatic expression (IE) processing and comprehension have challenged pre-trained language models (PTLMs) because their meanings are non-compositional. Unlike prior works that enable IE comprehension through fine-tuning PTLMs with sentences containing IEs, in this work, we construct IEKG, a commonsense knowledge graph for figurative interpretations of IEs. This extends the established ATOMIC2020 graph, converting PTLMs into knowledge models (KMs) that encode and infer commonsense knowledge related to IE use. Experiments show that various PTLMs can be converted into KMs with IEKG. We verify the quality of IEKG and the ability of the trained KMs with automatic and human evaluation. Through applications in natural language understanding, we show that a PTLM injected with knowledge from IEKG exhibits improved IE comprehension ability and can generalize to IEs unseen during training.