Translating Under Pressure: Domain-Aware LLMs for Crisis Communication

TL;DR

This paper presents a domain-adaptive translation pipeline that enhances crisis communication by expanding small datasets, fine-tuning language models, and biasing outputs toward simplified English for better readability during emergencies.

Contribution

It introduces a novel domain-aware approach combining data retrieval, filtering, fine-tuning, and preference optimization to improve crisis translation with limited data.

Findings

Improved readability and adequacy in crisis translation outputs.

Simplified English with domain adaptation serves as an effective lingua franca.

The approach outperforms baseline methods in automatic and human evaluations.

Abstract

Timely and reliable multilingual communication is critical during natural and human-induced disasters, but developing effective solutions for crisis communication is limited by the scarcity of curated parallel data. We propose a domain-adaptive pipeline that expands a small reference corpus, by retrieving and filtering data from general corpora. We use the resulting dataset to fine-tune a small language model for crisis-domain translation and then apply preference optimization to bias outputs toward CEFR A2-level English. Automatic and human evaluation shows that this approach improves readability, while maintaining strong adequacy. Our results indicate that simplified English, combined with domain adaptation, can function as a practical lingua franca for emergency communication when full multilingual coverage is not feasible.