大语言模型在汉英技术文献翻译中的应用实证研究  

Using Large Language Models for Chinese-English Technical Translation:An Empirical Study

作　　者：姚亚芝 YAO Yazhi(Beijing Jiaotong University)

机构地区：[1]北京交通大学

出　　处：《翻译界》2024年第2期1-17,共17页Translation Horizons

基　　金：教育部产学合作协同育人项目2022年第二批立项项目“产教研合作视域下的计算机辅助翻译课程案例教学研究与实践”(220905934225829)阶段性成果,合作单位:四川语言桥信息技术有限公司。

摘　　要：技术文献翻译对国际合作、知识共享和技术创新非常重要。本研究探索了大语言模型在汉英技术文献翻译中的应用,以提高翻译质量和效率。以铁路工程建设标准为例,使用GPT-4进行机器翻译和人机协同译后编辑,并对人机译文和人工译文进行多维度评估。结果显示,大语言模型在技术文献翻译中具有广泛的应用潜力,人机合作译文在语法、句法和逻辑方面优势显著,而人工译文在专业术语翻译上更准确。本研究提出了优化技术文献翻译流程的方法,强调术语表制作、提示优化等关键环节的重要性,并建立了基于大语言模型的译文质量评估体系,为科学评判技术文献翻译质量提供了新视角。The translation of technical literature plays a pivotal role in facilitating international collaboration,knowledge sharing,and technological innovation.This study investigates the application of large language models(LLMs)in Chinese-to-English technical document translation to improve translation quality and efficiency.Utilizing railway engineering construction standards as a case study,the GPT-4 model was employed for machine translation and human-machine collaborative post-editing,followed by multidimensional evaluations of both machine-assisted and human-translated texts.The findings reveal the extensive potential of LLMs in technical document translation,with machine-assisted translations exhibiting significant advantages in grammar,syntax,and logical coherence,while human-translated texts demonstrate superior accuracy in translating specialized terminology.Furthermore,the study proposes methods to optimize the technical document translation process,emphasizing the importance of term glossary preparation and prompt refinement,etc.By establishing a translation quality assessment framework based on LLMs,this research offers a novel perspective for scientifically evaluating the quality of technical document translation.

关 键 词：大语言模型 GPT-4 技术翻译 翻译质量 

分 类 号：H315.9[语言文字—英语]