Refining Translations with LLMs: A Constraint-Aware Iterative Prompting Approach

TL;DR

This paper introduces a multi-step, constraint-aware prompting method for LLMs that improves translation accuracy and robustness, especially in low-resource contexts, by integrating key term retrieval and iterative self-checking.

Contribution

The paper presents a novel iterative prompting approach that combines retrieval-augmented generation and self-refinement to enhance translation faithfulness in LLMs.

Findings

Significant improvements in translation accuracy on FLORES-200 and WMT datasets.

Enhanced robustness and faithfulness in low-resource translation scenarios.

Effective reduction of hallucinations through iterative self-checking.

Abstract

Large language models (LLMs) have demonstrated remarkable proficiency in machine translation (MT), even without specific training on the languages in question. However, translating rare words in low-resource or domain-specific contexts remains challenging for LLMs. To address this issue, we propose a multi-step prompt chain that enhances translation faithfulness by prioritizing key terms crucial for semantic accuracy. Our method first identifies these keywords and retrieves their translations from a bilingual dictionary, integrating them into the LLM's context using Retrieval-Augmented Generation (RAG). We further mitigate potential output hallucinations caused by long prompts through an iterative self-checking mechanism, where the LLM refines its translations based on lexical and semantic constraints. Experiments using Llama and Qwen as base models on the FLORES-200 and WMT datasets demonstrate significant improvements over baselines, highlighting the effectiveness of our approach in enhancing translation faithfulness and robustness, particularly in low-resource scenarios.