人工翻译与神经网络机器翻译译后编辑比较研究——基于对隐喻翻译的眼动追踪与键盘记录数据  

A comparative study of HT and NMT post-editing:data from eye-tracking and key-logging on metaphor translation

作　　者：王湘玲[1] 赖思 贾艳芳 WANG Xiangling;LAI Si;JIA Yanfang

机构地区：[1]湖南大学 [2]陕西国际商贸学院 [3]湖南师范大学

出　　处：《外语教学理论与实践》2021年第4期115-126,共12页Foreign Language Learning Theory And Practice

基　　金：国家社科基金一般项目“人工智能时代基于认知过程的翻译创新人才培养理论与实证研究”(19BYY104)的阶段性研究成果;湖南省社科基金重点项目“信息化时代高校翻译人才创新能力培养模式构建及实证研究”(17ZDB005)的支持。

摘　　要：本文通过眼动追踪、键盘记录、反省报告和问卷调查多元互证,比较了30名译者对隐喻表达进行人工翻译(HT)与译后编辑(PE)的认知负荷和译文质量。研究发现:无论是HT还是PE,译文区的注视时间均明显长于原文区,且隐喻译文的准确度均高于忠实度。不同之处含PE中原文区的注视次数、注视时间及单位单词停顿次数均显著少于HT,且隐喻译文的忠实度和准确度均显著高于HT。本研究表明PE在处理隐喻表达时,既可减少译者认知负荷还可提高译文质量。Based on triangulation of eye-tracking,key logging,retrospective protocols and questionnaires,the study aimed to compare post-editing with human translating metaphors from English into Chinese in terms of the cognitive effort of 30 translators and their translation quality.Results showed that for both human translation and post-editing,fixation duration was longer in TT AOIs than that in ST AOIs.Besides,the“accuracy”of metaphor translations was higher than“faithfulness”.Whereas in post-editing,the value of fixation counts,fixation duration in ST AOIs and TT AOIs,pause duration and PWR was all lower than that in human translation,despite the differences between fixation counts,fixation duration in TT AOIs and pause duration were not as significant.What’s more,post-editing produced significantly higher-quality metaphor translations than human translation in terms of“faithfulness”and“accuracy”.It can be reasonably concluded that post-editing not only reduced cognitive effort for English-Chinese metaphor translation but also improved the final quality.

关 键 词：神经网络机器翻译译后编辑 人工翻译 隐喻翻译 眼动追踪 键盘记录 

分 类 号：H085[语言文字—语言学] H059