本文是一篇旅游英语论文，本文从译员的角度出发，使用具有语言学意义的口译质量评价标准代替传统的机器翻译质量评估方法，探究当前机器口译App在旅游英语领域的口译质量，以及原语音质量、语音文字数量、男女语音差异与机器口译质量的关系。
Chapter One INTRODUCTION
1.1 Research Background
With the development of science and technology, people begin to pay more attention to the technology of interpretation in the era of artificial intelligence (Wang &Yang, 2019). Machine Interpretation, also known as “Spoken Language Translation (SLT)”, “speech-to-speech MT” or “Automated Interpretation”, refers to the process that the system translates speech from one language to another automatically (Rayner, Carter, Bouillon, Digalakis, & Wirén, 2003). MI includes three core technologies: speech recognition, machine translation, and speech synthesis. First, speech recognition, in which the machine receives the speech through software and then converts it into words; the second is the language conversion, that is, the machine processing and translating the text. The third is speech synthesis. That is, the machine will robotically form the target language then output.
Many well-known enterprises at home and abroad have launched the MI App, such as Baidu Translation, Google Translation, Sogou Translation, Youdao Translator, Microsoft Translation, Tencent Translator, etc. However, there are few empirical studies on MI, and it mainly focuses on automated and stereotypical scoring utilizing computer science algorithms and modeling, which is abstract and has no linguistic significance (B.H. Jiang, 2018; Liu, 2019).
1.2 Objectives and Significance
This thesis will supplement relevant evaluations and develop specific indicators suitable for MI in tourism, such as amplification and cultural transformation, pauses, and intonation of paralanguages, based on the quantitative and qualitative evaluation form of interpretation proposed by Yang (2005, p.237-238).
Furthermore, this thesis will simulate 9 scenes and 31 proposed requirements to collect eight subjects’ 248 sets of Chinese speech materials, according to the scene classification of “Travel English (3rd Edition)” (Y. Li, 2018) and “Travel with English alone” (Du, 2017).
After that, the “dialogue interpretation” of Youdao Translator will be used for Chinese-English interpretation which takes speech recognition, machine translation, and speech synthesis as a whole to conduct a comprehensive evaluation.
Finally, the quality of MI is evaluated, and suggestions for improvement are put forward.
Chapter Two LITERATURE REVIEW
2.1 The Development of MI
Since the middle and late 1980s, with the rapid development of automatic speech recognition (ASR) and machine translation (MT), machine interpretation has emerged. However, due to the lack of technologies, such as the internet and big data, MT is still groping. Since the limitations of the corpus, each MT system only supports a single field.
From 1993 to 2000, the Verbmobil, a hands-free dialogue translation system jointly developed by Germany, Japan, and the United States, could only handle German, English, and Japanese conversations in the business field. Since 2001, IBM has been conducting large-scale research on speech-to-speech translation. In September 2009, they launched ViaVoice Translator, a text-to-speech computer translation software, which laid the foundation for machine interpretation. Nevertheless, the project was eventually abandoned because there was no apparent progress.
Since 2011, with the maturity of speech recognition and machine translation, MI has developed rapidly, and its research has become a new hotspot in the field of information processing.
In 2012, Microsoft released a real-time English-Chinese MI system, which uses deep neural networks and statistical machine translation (SMT) to improve translation accuracy. In February 2013, Baidu released “Baidu Translator”, an App with speech recognition and conversation interpretation, becoming the first App with offline translation on the Android platform.
2
