Analysis Of Machine Translation Errors And Corresponding Pre-Editing And Post-Editing Strategies

The source text selected by this thesis is from The Human Brain Book,which systematically introduces the subject of human brain and neurology.This thesis takes the translation practice of this book as corpus,compares machine translation with human translation,classifies errors in the process of machine translation,analyzes the main causes of these errors and tries to put forward suggestions for modification,hoping to provide a reference for translators and other students.The author finds out that although machine translation based on neural network system has improved the translation efficiency and accuracy to a great extent,errors still exist.The author summarizes these errors into three categories.The first category is terminology translation error,including phrases of terminology and mistranslation of non-technical words.The second category is subject translation errors.When processing subjects,machine translation often makes mistakes,such as subject recognition errors,subject repetition,and subject positioning errors.The third category is logic error that is usually caused by conjunctions and parenthesis.For mistakes category 2 and 3,some sentence components like parenthesis and subordinate clause separate sentences into different sections and cause difficulties for machine translation.In this case,pre-editing could be applied to delete these sentence components from source text and add them to appropriate places in machine translation text in post-editing;for English sentence structures that contain pronouns,passive sentences,infinite verbs and gerunds which don't indicate the real subject of sentences or phrases and thus leading to wrong or vague translation results,pre-editing methods like complementing subjects to the source text can be helpful to improve post-editing efficiency and accuracy rate.At last,other kinds of mistakes made by machine translation include subject positioning errors,subject repetition errors and logic errors require translators to apply flexible post-editing strategies to correct machine translation text.