Post-editing Of Machine Translation: Errors Types And Strategies

Recently,translation demands related to e-commerce has drastically increased as the e-commerce flourished.Search queries,helping users to find the merchandise they want,is an essential part of e-commerce website.Nowadays,search queries translation gains more attentions as a new branch of website translation.With the purpose of presenting pragmatic strategies for search queries post-editing,this thesis,based on eBay search queries project,studies the text features of eBay search queries and the error types of machine translation output under the guidance of translation quality evaluation model.Through analysis,it is found that there are eight major forms of search queries.Some irregular forms of search queries make machine translation engine difficult to analyze them;besides,it also finds that machine translation engine is likely to make errors when translating search queries containing terms,proper nouns,colloquial words,and Chinese-featured words.With reference to DQF-MQM error categories and text features of search queries,mistranslation,redundancy,missing translation,and improper word order are identified as four major types of errors that need to be corrected in the machine translation output of eBay search queries.On the basis of post-editing practice,this thesis proposes seven strategies to post-edit machine translation output of search queries,that is,literal translation,free translation,application of annotations,application of parallel text,addition,omission,and adjustment of word order.Hopefully,it will improve the quality of search queries machine translation.