Figurative Language Understanding

This field studies how meaning changes when language is not meant literally, including metaphor, idiom, proverb, simile, and pun interpretation in text and images. Approaches range from contextual classifiers and span-level contrastive learning to linguistically guided annotation, semantic retrieval, knowledge graphs, and multimodal reasoning benchmarks. Recent work tests these methods across languages, cultures, conversational settings, and tasks such as detection, translation, explanation, and image grounding. Models can perform well on familiar patterns yet still favor literal readings, miss cultural cues, or give explanations unsupported by the input, especially for low-resource languages and multimodal content.