A Comparative Quality Assessment Of Google Online Translation And Human Translation Of Political Discourse From A CDA Perspective

The rapid development of end-to-end neural machine translation in recent years has drastically improved the quality of machine translation.In 2016,Google Neural Machine Translation(GNMT)made breakthroughs by bridging the gap between human and machine translation and enabling "zero-shot translation" through 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.Critical Discourse Analysis(CDA)researchers argue that language,power relations,and ideology interact with each other,and that language is not only influenced by social factors,but can also shape social realities.Therefore,the quality assessment of machine translation should go beyond the literal meaning comparison between the source language and the target language.The author of this thesis thus adopts a CDA perspective to evaluate human interpretation and the according translation of GNMT from a discourse level.The author uses the three-dimensional framework proposed by Norman Fairclough,one of the most influential critical discourse analysts.Through description,interpretation and explanation,the author studies how Foreign Minister Wang Yi responded to journalists'questions concerning the South China Sea and China's macro foreign policy in FMPRC's annual press conferences during the two sessions from 2015 to 2017.The author compares Sun Ning's interpretation with the translation of Google Translate,and through qualitative and quantitative analysis,the author assesses how the original information is delivered and recreated in both texts and whether and the intention of the speaker is achieved.Through comparative analysis,the author finds that Sun Ning's interpretation has higher lexical accuracy and closer logical cohesion than the corresponding translation of Google Translate.