An E-C Report On Cochrane Sinicization Project

With the rapid development of computer technology,machine translation + post-editing(MT+PE)has become a preferred mode for translators.To achieve high quality translation of such professional texts as medical texts,patents and contracts,machine translation currently cannot totally replace manual translation,and PE work is still indispensable.The author has participated in the Sinicization Project of the International Cochrane Evidence-based Medicine(EBM)Review from October 2020 to February 2022,during which the author post-edited thirteen pieces of EBM systematic review generated by the cloud MT platform,Memsource,a cloud-based translation management system specifically designed for Cochrane.Each task was accomplished with standardized procedures,going through machine translation,post-editing,first review,second review and final review by medical professionals.The final post-edited versions were all published on the Cochrane Collaboration website and the present report is based on the author’s MT+PE experience.During her translation practice,the author found that logical errors frequently occurred in Memsource MT output,which critically undermined the quality and readability of the translation.The author has focused on the MT logical errors,which are sorted into mistranslation of polysemy,wrong segmentation of logical relations,violation of general knowledge,mistranslation of logical connectives and translation inconsistency.She has used a simple statistical method to analyze the distribution and frequency of these logical errors.Inspired by the three principles of Skopos theory,the author has put forward the following suggestions for subsequent Memsource post-editing practice.Firstly,according to the Skopos rule,foreignization strategy can be adopted to enhance professional background knowledge and a technical glossary is beneficial for improving the accuracy of terminology translation,and cope with the mistranslation of logical connectives.Secondly,according to the coherence rule,specific context should be analyzed to correct the mistranslation of polysemous words,and information chunks should be integrated to redress the machine’s wrong segmentation of long and complex sentences.Finally,according to the fidelity rule,improper word order should be adjusted and logic relations should be verified through common sense,and grammatical analysis should be carried out to rectify over-translation and inconsistency.