Evaluation Of Google NMT's Translation Quality Based On Text Typology

At present,with the increasing intercultural communication and international competition,an increasing number of materials are needed to be translated.As translation costs numerous energy and time,machine translation comes into being.Machine translation has developed from the initial IBM model,to rule-based(RBMT)translation method and to phrase-based machine learning(PBMT).On September 27th,2016,Google Translate has introduced a new version of the translation program GNMT(Google neural machine translation system)which has caused great changes in the translation circle.GNMT is based on end-to-end automatic translation learning method.The accuracy rate of English-French and English-Spanish translation is over 90%.The accuracy rate of Chinese-English is about 80%.This paper is devoted to exploring the application of GNMT in the translation of political speeches,finding main problems in the translation of political speeches,and putting forward suggestions.This paper briefly reviews the history of machine translation and the problems in machine translation,the development of Google Translate,the feasibility and practicability of using GNMT for Chinese-English translation.Then,the paper discusses the text types,language features and the specific translation strategies.The paper analyzes the speeches in Xi Jinping:The Governance of China,classifies the text type of the speeches,and then makes a comparative analysis of the translation version by Google NMT and reference version from the aspects of vocabulary,syntax and text classification.Finally,the paper makes a conclusion that Google NMT can be applied to various aspects,but it cannot replace human translation.