Machine Translation And Post-editing Mode Based On DQF-MQM

With the development of artificial intelligence,machine translation(MT)related technologies have been continuously improved.However,due to the complexity of translation,there are still some mistakes like mistranslation and omission in MT output,so post-editing(PE)by the human translator is essential to guarantee the translation quality.Machine translation and post-editing(MT+PE)can leverage the advantages of both MT and human translation to achieve higher translation efficiency and better quality at the same time.MT+PE mode is also in innovation and progress.Although DQF-MQM has been widely used in language service industry,it has rarely been integrated into the workflow of MT+PE as a translation quality assessment(TQA)method.The major innovations of this translation practice lie in the organic combination of DQF-MQM and MT+PE and application of TQA before post-editing.Compared with the previous TQA metric,DQF-MQM contains a multidimensional and elaborated list of quality error typology,which can assist users in making a systematic and comprehensive analysis of MT errors.A comprehensive TQA report can decrease the cognitive load on translators and improve post-editing efficiency.The source text for this translation practice is selected from chapter 24 of Routledge Handbook of Translation and Technology,which is characterized by frequent and repeated use of terms and suitable for translation under MT+PE mode.The whole process of this translation practice is completed in SDL Trados 2021,a computer-aided translation software.With the help of Google Cloud Translation API,the author adopts Google Translate to get the first translation and then makes a TQA of MT output based on DQFMQM.After understanding the high-frequency errors of MT output,the author finally completes post-editing work,revision and language asset management.Based on the report of DQF-MQM,the author finds that errors in MT output are mainly from the accuracy and fluency dimensions,including the untranslated,mistranslation,under-translation,and errors of grammar and punctuations.After a quantitative analysis of the editing distance,the author presents the high-frequency errors of MT output and puts forward corresponding post-editing methods including verifying the accurate meaning of technical words,converting the voice of passive sentences,and adjusting the word order of long and complex sentences to highlight factual logic.Finally,the author systematically summarizes the focus and difficulties of translation work under MT+PE mode in the expectation of sharing the doings and findings with fellow translators using this mode.