Towards Sense-level Bilingual Dictionary Induction

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Expert, quick

Summary

A novel NLP task, Sense-Level Bilingual Dictionary Induction (SenseBDI), is introduced to automate the discovery of new bilingual sense entries, addressing the current manual, time-consuming, and subjective process of updating bilingual dictionaries. Existing methods like Word-level Bilingual Dictionary Induction and cross-lingual embedding alignment fail to account for polysemy or lexicographic data. The researchers constructed a time-stamped sense-level bilingual dictionary dataset by aligning two bilingual dictionaries, two monolingual dictionaries, and BabelNet, enriching bilingual entries with source-language details. Their proposed baseline, utilizing nearest-neighbor search over cross-lingual embeddings of glosses and usages, revealed that usages contribute more significantly than glosses, with performance varying across language pairs. The work also highlights challenges related to target language polysemy.

Key takeaway

For NLP Engineers developing cross-lingual lexical resources, this work suggests a shift towards sense-level induction to overcome polysemy limitations in traditional methods. You should prioritize integrating usage examples over mere glosses when building cross-lingual embeddings for sense alignment, as they offer stronger contributions. Consider the significant variation across language pairs, necessitating tailored approaches for optimal performance in real-world dictionary updates.

Key insights

Automating sense-level bilingual dictionary induction addresses polysemy gaps in lexicography using cross-lingual embeddings of usages and glosses.

Principles

• Bilingual dictionaries often lack completeness and diachronic information.
• Polysemy is a critical challenge for cross-lingual lexical tasks.
• Usages are more impactful than glosses for sense alignment.

Method

Construct a dataset by aligning two bilingual, two monolingual dictionaries, and BabelNet. Apply nearest-neighbor search on cross-lingual embeddings of glosses and usages.

In practice

• Enrich bilingual entries with monolingual source-language data.
• Prioritize usage examples over glosses for sense alignment.
• Consider language-pair specific variations in embedding performance.

Topics

• Sense-Level Bilingual Dictionary Induction
• Cross-lingual Embeddings
• Lexical Semantics
• Polysemy
• Bilingual Dictionaries
• BabelNet

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• Generating Concept Lexicalizations via Dictionary-Based Cross-Lingual Sense Projection — A dictionary-augmented projection method improves cross-lingual WordNet expansion precision with minimal resources.
• When Similar Means Different: Evaluating LLMs on Arabic--Hebrew Cognates — Current LLMs struggle with cross-lingual semantic disambiguation, especially false friends and loanwords, due to over-reliance on surface forms.
• IHUBERT: Vector-Based Semantic Deduplication and Domain-Balanced Pretraining for Persian Resources — Semantic deduplication and domain-balanced pretraining significantly enhance Persian PLM performance across diverse NLU tasks.
• G-IdiomAlign: A Gloss-Pivoted Benchmark for Cross-Lingual Idiom Alignment — Cross-lingual idiom alignment challenges LLMs due to non-compositionality, with glosses offering semantic grounding but literal bias remaining dominant.
• Global vs. Local Sentence Embeddings for Brazilian Portuguese: Revisiting Monolingual Models in the Age of Foundation Models — Optimal model choice for PT-BR sentence embeddings balances vocabulary coverage and adaptation flexibility.
• From TF-IDF to Transformers: Implementing Four Generations of Semantic Search — Semantic search evolved from explicit rules to learned contextual representations, balancing interpretability with semantic depth.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.