Idioms Share Conceptual Roots Across Languages

Researchers developed a network framework to analyze 160 idiomatic expressions from eight diverse languages. They annotated each expression with conceptual features, such as containment, emotional impact, or social context, based on cognitive-linguistic theory.

The resulting network revealed that idioms group by underlying conceptual schema rather than language of origin, supporting existing theories about how humans understand figurative language. The framework goes beyond simple word associations.

It captures semantic information that standard computational linguistics methods miss and can be expanded using large language models. Experiments showed the network accurately identifies potential translation equivalents between five language families. The team found that all conceptual feature dimensions contribute uniquely to the network’s structure and its ability to detect idioms, suggesting a robust and interpretable system for understanding cross-linguistic meaning.