Statistics show that artificial translation output for average is 2000 words per person per 8 hours; however, translation system can do 3000 characters per second. Computer translation system can complete 86400000 words in the same 8 hours, which
is 43200 times as fast as artificial translation (Li Qingzhao，303-307). It is obviously that the MT has a great edge over human translation in higher translation speed and lower cost.
However，Natural language is complex. The same sentence may have the different meanings under different circumstances. This complexity is made of the rational and irrational elements. It is very difficult for MT to match the appropriate contexts reflected in the target languages because computer has no cognitive and comprehensive ability as man. So in many cases, especially in political text, the results of MT are of poor quality.
2. Literature Review
2.1 Study on machine translation
European Association of Machine Translation defined Machine Translation as “ the process of using computer to translate a natural source language into another natural target language. In general, it refers to sentences and full text translation of a natural language.” The whole process of translation has been completed all by the computer without artificial participation, including analyzing structures of a source language, and converting corresponding vocabulary and structures to a target language.
In 1954, America Georgetown University and International Business Machines Corporation (IBM) first carried joint experiments using computer machine translation system, which realized the first translation system beyond the word to word concept, marking the birth of the real Machine Translation system.
Since then, the development of MT experienced ups and downs. Nowadays, MT technology has fruitful achievements such as “Google online” and “Youdao online (Netease)”. They’ve used Neural Machine Translation Technology and made the translation quality and efficiency improved significantly. Neural Machine Translation (NMT) has recently been introduced as a promising approach with the potential of addressing many shortcomings of traditional machine translation systems. According to Google’s report, the strength of NMT lies in its ability to learn directly, in an end-to-end fashion, the mapping from input text to associated output text. Its architecture typically consists of two recurrent neural networks (RNNs), one to consume the input text sequence and one to generate translated output text. NMT is often accompanied by an attention mechanism (Bahdanau et al. In International Conference on Learning Representations (2015) ) that helps it cope effectively with long input sequences.
The latest achievement of MT is Neural Machine Translation (NMT). NMT is an end-to-end learning approach for automated translation, with the potential to overcome many of the weakness of conventional phrase-based translation system. In 2016 Oct.8th,
Google’s engineers presented GNMT, Google’s Neural Machine Translation system, which attempts to address many of these issues. GNMT model consists of a deep LSTM network with 8 encoder and 8 decoder layers using residual connections as well as attention connections from the decoder network to the encoder（Yonghui Wu et al. 1）.
This thesis will choose Google Neural Machine Translation as an outstanding representative of MT to compare GNMT’s translation quality with human translation. This thesis will pick translator and GNMT’s different translation of political text as research object. To be more specific, it will choose 2018’s Report on the Work of the Government delivered at the First Session of the 13th National People’s Congress of the People’s Republic of China on March 5th, 2018 as research text. Referring to Julian House’s translation quality assessment, this thesis is to evaluate GNMT’s translation quality of political text and to discuss the limitation of GNMT and MT faced nowadays. Most importantly, this thesis will conclude whether GNMT can take on the translation of political translation. 对比分析人工翻译与谷歌神经网络翻译对于政治文本的翻译质量(2):http://www.chuibin.com/yingyu/lunwen_206644.html