문학작품 기계번역의 한계에 대한 상세 고찰

Abstract

Mah, Seung-Hye. (2018). A Detailed Investigation on Limitations of Literary Work Machine Translation. Interpreting and Translation Studies 22(3) , 65-88 Artificial Intelligence(AI) is all around us, from phones and cars to even conducting surgery and writing novels, impacting our everyday lives and work. The field of interpreting and translation has also been daunted by the rapid development of machine interpreting and translation. Certain parts of the interpreting and translation processes would surely been replaced by machine working; however, other parts, especially those that require delicate transfer of feelings and contexts, will remain in the domain of human working. Literary translation is a good case in point. Literary translation from Korean into English, in particular, needs to reflect the differences in the usage of both languages. Deborah Smith, the translator of The Vegetarian , mentioned that she had paid a lot of attention to transfer Korean that has characteristics of ambiguity, repetition, and plain prose to English that is characterized by precision, concision, and dynamism. She harnessed her competences of interpretation, decision-making, creative representation, and communication for readers to translate The Vegetarian , reflecting the differences between Korean and English. On the contrary, Google, one of the most frequently used neural machine translation systems, does not recognize the differences between Korean and English, and the differences are not reflected in literary translation, which gives rise to critical errors in final translation outcome. This detailed investigation on errors in literary translations done by Google gives us implications as to what machine translation cannot take over and where human translators can stand.