【作者】 李逢泽；

【导师】 王建军；

【作者基本信息】 内蒙古大学 ， 翻译硕士（专业学位）， 2023， 硕士

【摘要】 伴随计算机科学、神经网络和仿生学近年间突飞猛进的发展,人工智能也在逐渐进入人们的视野和日常生活。语音识别作为人工智能方向的一个重要分支,在诸多领域有了实质性的进展,从语音转文字到智能助手和智能家电,甚至还应用到口译领域。然而实践证明,单一的基于数据库的机器翻译结果始终不能令人完全满意,但该技术的语音识别环节已趋于成熟。基于丹尼尔·吉尔的认知负荷理论,本文将探讨以语音识别技术代替交替传译中口译员的笔记工作的可行性。本文将选用目前国内语音识别领域最为出色的科大讯飞公司旗下的讯飞听见作为实验工具,并以对照实验的形式探索语音识别对于口译员的效率性影响。实验选择了10名内蒙古大学的英语口译研究生作为实验对象,并依据多种不同变量做出对比。为使数据更为全面,实验分为汉译英和英译汉两组。在客观数据之外,实验对象还完成了实验问卷和实验反馈,以此达到最理想的实验结果分析。经过多次对照实验,可得出结论:在英译汉语音识别准确率为98.21%时,译员译文评分平均提高10.5%;在汉译英语音识别准确率为97.01%时,译员译文评分平均提高6.2%。在忠实度、表达度、语言性的分支评判标准中,每个译员的表现各不相同,但在时间控制方面,大部分译员产出译文花费的时间都要比原文更多。综合上述内容,实验结果表明,在交替传译过程中,语音识别总体而言可以为译员能够带来正面作用,但作用的大小因人、因材料难度而异。更多还原

【Abstract】 With the rapid development of computer science,neural networks,and bionics in recent years,artificial intelligence is gradually entering people’s vision and daily life.As an important branch of artificial intelligence,speech recognition has made substantial progress in many fields,from speech-to-text service to intelligent assistants and smart home,it even applied to the field of interpretation.However,practice has proven that database-based machine translation result is usually not satisfactory.Different from machine interpretation,the speech recognition technology has become mature already.Based on Daniel Gill’s cognitive load theory,this article will explore the feasibility of using speech recognition technology to replace the note-taking work of interpreters in consecutive interpretation.In this thesis,the i FLYREC,a sub-product of i FLYTEK,which is currently the most outstanding in the field of speech recognition in China is used as the experimental tool.And the main task of this thesis is to explore the impact of speech recognition on the efficiency of interpreters by conducting sets of comparative experiments.In the experiments,10 graduate students of English interpretation from Inner Mongolia University were selected as the subjects,and comparisons were made based on several different variables.In order to make the data more comprehensive,the experiment was divided into two groups: Chinese-English interpretation and English-Chinese interpretation.In addition to objective data,experimental subjects also completed questionnaires and feedback to achieve the optimal analysis of the results.After multiple controlled experiments,it can be concluded that when the accuracy rate of English-Chinese speech recognition is 98.21%,the interpreters’ score increases by an average of 10.5%;When the accuracy rate of Chinese-English speech recognition is97.01%,the score of translators’ interpreting results increases by an average of 6.2%.In terms of criteria of Fidelity,Delivery,and Language,each translator’s performance varies,but in terms of time control,the duration of most translators’ interpreting results is more than the original.With these data taken into account,the experimental results show that in the process of consecutive interpretation,speech recognition can generally bring positive effects to the interpreter,but the magnitude of the effects varies depending on the person and the difficulty of the material.更多还原