Towards Style Alignment in Cross-Cultural Translation

2025/06/30 by Havaldar, Shreya, Stein, Adam, Wong, Eric +1 · 3 citations
#Computation and Language (cs.CL) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2507.00216

Abstract

Successful communication depends on the speaker's intended style (i.e., what the speaker is trying to convey) aligning with the listener's interpreted style (i.e., what the listener perceives). However, cultural differences often lead to misalignment between the two; for example, politeness is often lost in translation. We characterize the ways that LLMs fail to translate style - biasing translations towards neutrality and performing worse in non-Western languages. We mitigate these failures with RASTA (Retrieval-Augmented STylistic Alignment), a method that leverages learned stylistic concepts to encourage LLM translation to appropriately convey cultural communication norms and align style.

Citations

• Retrieval-Augmented Machine Translation with Unstructured Knowledge
• Towards Cross-Cultural Machine Translation with Retrieval-Augmented Generation from Multilingual Knowledge Graphs
• The Llama 3 Herd of Models
• Multilingual Text Style Transfer: Datasets & Models for Indian Languages
• M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation
• Translate Meanings, Not Just Words: IdiomKB's Role in Optimizing Idiomatic Translation with Language Models
• Benchmarking Machine Translation with Cultural Awareness
• QLoRA: Efficient Finetuning of Quantized LLMs
• Large Language Models Are State-of-the-Art Evaluators of Translation Quality
• The unreasonable effectiveness of few-shot learning for machine translation
• In-context Examples Selection for Machine Translation
• TyDiP: A Dataset for Politeness Classification in Nine Typologically Diverse Languages
• SemEval 2023 Task 9: Multilingual Tweet Intimacy Analysis
• FRMT: A Benchmark for Few-Shot Region-Aware Machine Translation
• CometKiwi: IST-Unbabel 2022 Submission for the Quality Estimation Shared Task
• Controlling Translation Formality Using Pre-trained Multilingual Language Models
• CoCoA-MT: A Dataset and Benchmark for Contrastive Controlled MT with Application to Formality
• Challenges and Strategies in Cross-Cultural NLP
• Few-shot Controllable Style Transfer for Low-Resource Multilingual Settings
• Towards Universality in Multilingual Text Rewriting
• XFORMAL: A Benchmark for Multilingual Formality Style Transfer
• TextSETTR: Few-Shot Text Style Extraction and Tunable Targeted Restyling
• Unsupervised Cross-lingual Representation Learning at Scale
• Multi-Task Neural Models for Translating Between Styles Within and Across Languages
• Dear Sir or Madam, May I introduce the GYAFC Dataset: Corpus, Benchmarks and Metrics for Formality Style Transfer
• Cross-Cultural Pragmatic Failure

Cited by

• Semantic Label Drift in Cross-Cultural Translation
• Culturally-Aware Conversations: A Framework & Benchmark for LLMs
• VideoNorms: Benchmarking Cultural Awareness of Video Language Models

Related

• Designing LLMs for cultural sensitivity: Evidence from English-Japanese translation
• Rethinking Cross-lingual Alignment: Balancing Transfer and Cultural Erasure in Multilingual LLMs
• Comparing Styles across Languages: A Cross-Cultural Exploration of Politeness
• Semantic Label Drift in Cross-Cultural Translation
• Cultural Alignment in Large Language Models: An Explanatory Analysis Based on Hofstede's Cultural Dimensions