When Words Don't Mean What They Say: Figurative Understanding in Bengali Idioms

TL;DR

This paper introduces a comprehensive Bengali idiom dataset and benchmark to evaluate and improve large language models' understanding of figurative language in low-resource, culturally-rich contexts, revealing significant performance gaps.

Contribution

The creation of a large-scale, culturally-grounded Bengali idiom dataset with detailed annotations and a benchmark for evaluating LLMs' figurative language understanding.

Findings

No model exceeds 50% accuracy on the benchmark.

Human performance is 83.4%, highlighting the models' limitations.

The dataset and benchmark reveal significant gaps in LLMs' cultural and figurative understanding.

Abstract

Figurative language understanding remains a significant challenge for Large Language Models (LLMs), especially for low-resource languages. To address this, we introduce a new idiom dataset, a large-scale, culturally-grounded corpus of 10,361 Bengali idioms. Each idiom is annotated under a comprehensive 19-field schema, established and refined through a deliberative expert consensus process, that captures its semantic, syntactic, cultural, and religious dimensions, providing a rich, structured resource for computational linguistics. To establish a robust benchmark for Bangla figurative language understanding, we evaluate 30 state-of-the-art multilingual and instruction-tuned LLMs on the task of inferring figurative meaning. Our results reveal a critical performance gap, with no model surpassing 50% accuracy, a stark contrast to significantly higher human performance (83.4%). This underscores the limitations of existing models in cross-linguistic and cultural reasoning. By releasing the new idiom dataset and benchmark, we provide foundational infrastructure for advancing figurative language understanding and cultural grounding in LLMs for Bengali and other low-resource languages.