Contextualising (Im)plausible Events Triggers Figurative Language
This work addresses the challenge of understanding figurative language processing for NLP researchers, but it is incremental as it builds on existing plausibility and literalness studies.
The study investigated how humans and large language models (LLMs) differ in assessing plausibility and non-literalness in English event triples, finding that humans excel at nuanced detection and contextualization, while LLMs show shallow patterns and a bias toward interpreting implausible events as non-literal.
This work explores the connection between (non-)literalness and plausibility at the example of subject-verb-object events in English. We design a systematic setup of plausible and implausible event triples in combination with abstract and concrete constituent categories. Our analysis of human and LLM-generated judgments and example contexts reveals substantial differences between assessments of plausibility. While humans excel at nuanced detection and contextualization of (non-)literal vs. implausible events, LLM results reveal only shallow contextualization patterns with a bias to trade implausibility for non-literal, plausible interpretations.