BridG MT: Enhancing LLMs' Machine Translation Capabilities with Sentence Bridging and Gradual MT

2024-10-15

AI Summary

Large language models (LLMs) exhibit limited performance in machine translation for low-resource languages, and existing few-shot approaches rely heavily on high-quality external parallel examples—often unavailable in such settings. Method: This paper proposes a lightweight, zero-external-parallel-corpus enhancement framework centered on a novel “sentence bridging” and “progressive translation” co-mechanism: it generates semantically coherent intermediate sentence sequences to gradually map source to target sentences, while dynamically constructing in-context examples from early-stage translations. The method integrates prompt engineering, chain-of-translation reasoning, and dynamic context construction, and is compatible with mainstream black-box LLMs. Results: Experiments across four LLMs and seven low-resource languages demonstrate that our approach significantly outperforms conventional few-shot baselines requiring extensive external examples, achieving more robust and efficient zero-shot and few-shot translation without additional parallel data.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Recent Large Language Models (LLMs) have demonstrated impressive translation performance without requiring fine-tuning on additional parallel corpora. However, they still face significant challenges in certain scenarios, particularly when translating low-resource languages. A common approach to address this issue is to provide external knowledge, such as few-shot examples, to assist LLMs in translating specific source sentences. However, this method is fundamentally limited by the quality or quantity of relevant sources, which cannot always be guaranteed. To reduce LLMs' reliance on external sources, we propose BridG MT, a method that combines Sentence Bridging, which generates a sequence of sentences as a bridge that gradually transition from easy-to-translate to more difficult, and Gradual MT, which sequentially translates these sentences using earlier translations as few-shot examples for subsequent ones. Experiments conducted on four LLMs across seven languages demonstrate that our method effectively enhances translation performance, even outperforming translation methods that rely on a large number of few-shot examples.

Problem

• Enhancing LLM translation for low-resource languages
• Reducing reliance on external knowledge sources
• Improving translation without fine-tuning parallel corpora

Innovation

• Sentence Bridging generates gradual transition sentences
• Gradual MT uses prior translations as few-shot examples
• Reduces reliance on external knowledge sources