Conceptual Networks for Cross-Linguistic Idiomatic Expressions:A Feature-Based Graph Approach

2026-07-10 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

A new interpretable network-based framework represents idiomatic and figurative meaning across eight typologically diverse languages, encompassing 160 conventional expressions. This framework annotates each expression with binary conceptual features, derived from cognitive-linguistic theory, and uses pairwise Jaccard similarities to construct a weighted graph. Community detection within this graph reveals that idioms cluster by conceptual schema rather than by language, aligning with cognitive-linguistic predictions. The conceptual network captures unique semantic information distinct from distributional embeddings, can be scaled through automatic annotation with LLMs, and enhances downstream idiom detection. It also remains robust when enriched with corpus frequencies. Cross-lingual transfer experiments demonstrate that conceptual proximity alone can identify acceptable translation equivalents across five language families, achieving substantial gains over embedding-based baselines. Ablation studies confirm that all three feature dimensions—schemas, roles, and valence—contribute non-redundantly to the network's organization and idiom detection performance.

Key takeaway

For NLP Engineers developing cross-lingual systems, this framework offers a robust approach to handling idiomatic expressions. You should consider integrating conceptual network analysis, particularly for tasks like idiom detection and translation equivalent identification, as it outperforms embedding-based baselines. Leveraging automatic annotation with LLMs can scale this method, providing interpretable, stable representations of idiomatic meaning across diverse languages.

Key insights

Idioms cluster by conceptual schema, not language, enabling cross-lingual transfer and improved detection.

Principles

• Idioms cluster by conceptual schema, not language.
• Conceptual networks capture unique semantic information.
• Feature dimensions (schemas, roles, valence) are non-redundant.

Method

Annotate expressions with binary conceptual features, define a weighted graph using Jaccard similarities, then apply community detection.

In practice

• Scale annotation with LLMs for broader application.
• Improve idiom detection using graph-derived signals.
• Identify cross-lingual translation equivalents via conceptual proximity.

Topics

• Conceptual Networks
• Idiomatic Expressions
• Cross-Lingual Transfer
• Graph-Based Methods
• LLM Annotation
• Cognitive Linguistics

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• Conceptual Networks for Cross-Linguistic Idiomatic Expressions:A Feature-Based Graph Approach — Idioms organize by conceptual schema, not language, offering interpretable, cross-linguistically stable meaning representations.
• Discovering Lexical Gaps Using Embeddings from Multilingual LLMs — Cross-lingual lexical gaps can be identified by measuring semantic alignment weakness in multilingual LLM embeddings.
• G-IdiomAlign: A Gloss-Pivoted Benchmark for Cross-Lingual Idiom Alignment — Cross-lingual idiom alignment in LLMs benefits from explicit semantic glosses, yet literal translation bias remains a significant challenge.
• Language-Specific versus Cross-Lingual Knowledge Graphs for Implicit Aspect Identification in Arabic: A Comparative Study of Reasoning and Adaptation Strategies — Native Arabic knowledge graphs and task-specific fine-tuning are crucial for effective implicit aspect identification in Arabic ABSA.
• Construction of Knowledge Graph based on Language Model — PLMs significantly enhance Knowledge Graph construction by automating entity and relation extraction from text.
• Universal Conceptual Structure in Neural Translation: Probing NLLB-200's Multilingual Geometry — NLLB-200's internal geometry reflects language-universal conceptual structures, akin to human multilingual cognition.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.