A Discourse Translation Perspective On AI Translation

The thesis takes a glimpse at Chinese-English(C-E)AI translation from the perspective of discourse translation on the basis of both theoretical analysis and discourse analysis.The materials from C-E Translation Course of Graduate School of Translation and Interpretation(GSTI)of Beijing Foreign Studies University are used as source language data,while the translation generated by Google Neural Machine Translation(GNMT)and the translation made by the teachers team are used as AI-generated target language data and human-generated target language data respectively.The discourse analysis has been carried out from the perspectives of five cohesive devices,namely,reference,substitution,ellipsis,conjunction,and lexical cohesion.By the comparison and contrast,the thesis aims to figure out specific causes of AI translation’s poor performance in regard to discourse translation,so as to inspire translation students.As for reference,AI translation may wrongly generate the reference when it comes to historical and cultural background,extended meanings,and semantic fuzziness.As for substitution,AI translation has defects in translation independency and the economical principle of language.As for ellipsis,AI translation differs from human translation mainly in message representation,initiative ellipsis,and punctuation marks utilization.As for conjunction,AI translation shows its shortcomings in judging natural order and decoding implicit logic relations.As for lexical cohesion,AI translation is about technology while human translation is about art.