机器口译软件翻译质量评估实验研究——以中英旅游文本为例

An Experimental Study on Translation Quality of Machine Interpretation App:Taking China-UK Tourism Texts as Example

近年来机器口译技术发展迅速,多种口译软件应时而生,但其翻译质量如何,有待结合具体应用场景进行评测。本文选取"有道翻译官"口译软件,模拟中英旅游口译场景,以8位测试者248组对话为实验材料,对机器口译质量进行量化与质化评测,提出了面向旅游翻译的机器口译质量评估标准。研究发现,当前机器口译App在旅游方面的应用尚未达到预期效果,在语音识别、文化转化、冗余删减、称谓名称处理、数字和语音效果等6个方面尚需不断改进。此外,还发现原语音质量与机器口译正相关,而语音文字数量与机器口译质量负相关,男女语音差异不会对机器口译质量产生明显影响。

With the rapid development of machine interpretation(MI),a variety of related Apps have made great progress,but interpretation quality of these Apps needs to be evaluated in the specific scenarios.Based on the tourism dialogue interpreting,this paper simulated the real scene,248 conversations from 8 subjects as experiment materials,Youdao Translator as the tool to carry out a quantitative and qualitative experimental study on the quality of MI.On the whole,the current application of MI Apps in tourism has not achieved expected results,and MI Apps still need to be improved in six aspects,including speech recognition,cultural transformation,redundancy reduction,appellation and item name,number and phonetic effect.The experiments found that the higher the quality of the original voice,the higher the quality of MI,and the higher the number of spoken words,the lower the quality.Besides,the difference between male and female voices will not have a significant impact on the quality of MI.