摘要
机器翻译在信息型文本翻译中接受度良好,但其应用于创造性的文本翻译时存在争议。本文聚焦文学文本的机器翻译译后编辑,在词汇与句子层面对比了学生译者在神经机器翻译译后编辑和人工翻译模式下的汉语小说英译译文。研究发现,在词汇层面,人工翻译比译后编辑采用了更为多样的词汇解决方案,词语翻译熵更高,译者的创造性空间更大,机器翻译译文对译后编辑具有语义上的引导和限制。在句子层面,两种模式下的译文均较原文缩短了平均句长,译后编辑比人工翻译拆译更多、合译更少,与机器翻译输出的拆译频率接近。最后,从译后编辑认知过程的角度对结果进行解释,探讨研究结果对翻译技术教育的启示。
Although machine translation has been well-accepted in the translation practice of informative texts,its application to creative text translation remains controversial.The present study focuses on the post-editing of literary texts translated by neural machine translation.It compares student translators'English translations of Chinese novels in two translation modalities:Neural machine translation post-editing and from-scratch human translation,analyzing the translated texts at lexical and syntactic levels.The findings reveal that at the lexical level,human translation presents more diverse lexical solutions and exhibits higher lexical translation entropy,offering greater creative freedom for translators.In contrast,machine-translation outputs semantically constrained post-editing.At the sentence level,translations in both modalities shortened the average sentence length compared with the source texts.However,post-editing involves more sentence splitting and fewer merging operations than human translation,with the frequency of splitting closely aligned with that of the machinetranslation output.The study interprets the results from the perspective of the cognitive process of post-editing and discusses their implications for translation technology education.
出处
《外语与翻译》
2025年第1期1-9,I0001,共10页
Foreign Languages and Translation
基金
广东省哲学社会科学规划外语信息化专项一般项目“基于三元互证模式的翻译方向性研究”(项目号:GD23WZXC02-05)
广东省教育科学规划课题(高等教育专项)“新发展格局下粤港澳大湾区多语种人才自主培养体系研究”(项目号:2024GXJK593)
广东外语外贸大学研究生科研创新项目“神经机译辅助下的文学文本译后编辑语言特征研究”(项目号:24GWCXXM-086)的阶段性成果。
作者简介
卢植,广东外语外贸大学高级翻译学院教授/华侨大学外国语学院教授;李珂,广东外语外贸大学高级翻译学院博士生,通信地址:510420广东省广州市白云区白云大道北2号广东外语外贸大学高级翻译学院。
