arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2412.05862cs.CL
收藏

基于开源大语言模型的领域特定翻译:面向资源的分析

Domain-Specific Translation with Open-Source Large Language Models: Resource-Oriented Analysis

Aman Kassahun Wassie(无);Mahdi Molaei(无);Yasmin Moslem(无)

Aman Kassahun Wassie, Mahdi Molaei, Yasmin Moslem

AI总结:

本研究比较了开源大语言模型与专用机器翻译模型在医学领域多资源条件下的翻译性能,发现微调后的 NLLB-200 3.3B 仍优于 7-8B 参数的 LLMs,强调了专用 MT 模型在低资源翻译中的必要性及预训练中型领域模型的价值。

摘要:

在这项工作中,我们比较了开源自回归仅解码器大语言模型(LLMs)与面向任务的机器翻译(MT)模型在领域特定翻译中的性能。我们的实验聚焦于医学领域,涵盖四种资源可用性不同的语言方向:英语到法语、英语到葡萄牙语、英语到斯瓦希里语以及斯瓦希里语到英语。尽管近期取得了进展,但与 NLLB-200 等多语言编码器-解码器 MT 模型相比,LLMs 在专业翻译中仍表现出显著的质量差距。结果表明,在四个语言方向中的三个方向上,NLLB-200 3.3B 的表现优于所有评估的 7-8B 参数范围内的 LLMs。虽然微调提升了 Mistral 和 Llama 等 LLMs 的性能,但与微调后的 NLLB-200 3.3B 模型相比,这些模型的表现依然不佳。我们的发现强调,为了实现高质量的领域特定翻译,仍然需要专门的 MT 模型,尤其是在中等资源和低资源环境下。此外,较大 LLMs 相比其 8B 变体表现更优,这表明预训练领域特定的中型语言模型具有潜在价值,采用针对性的数据选择和知识蒸馏方法可以提升专业翻译任务的质量与效率。

英文摘要:

In this work, we compare the domain-specific translation performance of open-source autoregressive decoder-only large language models (LLMs) with task-oriented machine translation (MT) models. Our experiments focus on the medical domain and cover four language directions with varied resource availability: English-to-French, English-to-Portuguese, English-to-Swahili, and Swahili-to-English. Despite recent advancements, LLMs demonstrate a significant quality gap in specialized translation compared to multilingual encoder-decoder MT models such as NLLB-200. Our results indicate that NLLB-200 3.3B outperforms all evaluated LLMs in the 7-8B parameter range across three out of the four language directions. While fine-tuning improves the performance of LLMs such as Mistral and Llama, these models still underperform compared to fine-tuned NLLB-200 3.3B models. Our findings highlight the ongoing need for specialized MT models to achieve high-quality domain-specific translation, especially in medium-resource and low-resource settings. Moreover, the superior performance of larger LLMs over their 8B variants suggests potential value in pre-training domain-specific medium-sized language models, employing targeted data selection and knowledge distillation approaches to enhance both quality and efficiency in specialized translation tasks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑