Improving LLM-based Document-level Machine Translation with Multi-Knowledge Fusion

2025-03-15

AI Summary

Document-level machine translation (DMT) suffers from limited contextual modeling and poor long-range coherence due to overreliance on raw sentence sequences. Method: This paper proposes a dual-knowledge-source-enhanced large language model (LLM) framework that jointly incorporates document summarization and entity translation as parallel knowledge sources. It employs a two-stage generation process coupled with multi-knowledge collaborative re-ranking, alleviating the bottleneck of single-sequence context dependency. The approach integrates prompt engineering, knowledge distillation, and COMET-guided dynamic filtering and fusion ranking. Contribution/Results: Evaluated on LLaMA3-8B-Instruct, Mistral-Nemo-Instruct, and GPT-4o-mini, the method achieves average COMET score improvements of +0.8, +0.6, and +0.4 across eight DMT benchmarks, respectively—significantly outperforming knowledge-free baselines. Results validate that dual-knowledge co-modeling enhances both translation consistency and accuracy.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• 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

Abstract

Recent studies in prompting large language model (LLM) for document-level machine translation (DMT) primarily focus on the inter-sentence context by flatting the source document into a long sequence. This approach relies solely on the sequence of sentences within the document. However, the complexity of document-level sequences is greater than that of shorter sentence-level sequences, which may limit LLM's ability in DMT when only this single-source knowledge is used. In this paper, we propose an enhanced approach by incorporating multiple sources of knowledge, including both the document summarization and entity translation, to enhance the performance of LLM-based DMT. Given a source document, we first obtain its summarization and translation of entities via LLM as the additional knowledge. We then utilize LLMs to generate two translations of the source document by fusing these two single knowledge sources, respectively. Finally, recognizing that different sources of knowledge may aid or hinder the translation of different sentences, we refine and rank the translations by leveraging a multi-knowledge fusion strategy to ensure the best results. Experimental results in eight document-level translation tasks show that our approach achieves an average improvement of 0.8, 0.6, and 0.4 COMET scores over the baseline without extra knowledge for LLaMA3-8B-Instruct, Mistral-Nemo-Instruct, and GPT-4o-mini, respectively.

Problem

• Enhance document-level machine translation using multi-knowledge fusion.
• Incorporate document summarization and entity translation for better LLM performance.
• Refine translations by ranking with a multi-knowledge fusion strategy.

Innovation

• Incorporates document summarization and entity translation
• Fuses multiple knowledge sources for translation
• Refines translations using multi-knowledge fusion strategy