Translation Quality Assessment Of DeepL Translator For Political Texts Based On Multidimensional Quality Metrics

The continuing exponential growth of diverse translation tools in recent years necessitates further requirements for Translation Quality Assessment(TQA)in academic,institutional and industry settings.The operability of TQA models is always the focus and challenge in TQA research,and thus new standard error typologies,introduced to replace the previous TQA models,are being taken on.With this in mind,Multidimensional Quality Metrics(MQM),based on the Error Analysis theory,summarizes 145 issue types in all and 19 ones at core,by integrating all error types proposed by LISA QA Model,SAE J2450,etc.In the long-term practice,MQM has been proved to be the favored TQA model by researchers that falls under holistic and analytical approaches.It means that these error categories can be used to assess the text as a whole or on a sentence-by-sentence basis as in this study.Besides,the framework proposes a variety of error categories that can be drawn upon to create customized metrics based on the needs of the end user.Given the above context,this research attempts to assess the translation quality of DeepL Translator,a new translation tool entering into the Chinese market,by employing MQM as the TQA model.The Chinese version of Report on the Work of the Government(2020)is selected from the official website since its authority and reliability;and its official human translation version and the output of DeepL Translator are compared for the analysis in the research.The statistical results show that DeepL Translator,compared with the human translation,does not have a satisfactory performance as expected in the dimensions of Accuracy and Fluency,especially in terms of Grammar and Terminology.