This study aims to identify the limitations and availability of neural machine translation(NMT) through a comparative analysis of the Arabic-Korean human translation and NMT of a literary text. To this end, the NMT of Saud Alsanousi’s novel The Bamboo Stalk is analyzed compared to its human translation published in Korean. Errors in the NMT output are classified into nine categories for quantitative analysis, and exemplary errors are qualitatively examined. According to the findings of analysis, many errors involve polysemous words that must be translated according to their context. In particular, when negative words are included, the whole sentence is often translated into the opposite meaning. The explicitation of pronouns has not been done as much as predicted. The Arabic phrases of inanimate subject with animate object appeared and the inanimate subjects of Arabic were rendered into other sentence components. Specifically, they were rendered into adverb phrases and predicates. In addition, this study presents good cases of NMT to confirm it’s availability. Although this study uses limited data for analysis, it is meaningful in that it confirms the limitations and availability of Arabic-Korean NMT.
Ⅰ. 서론
Ⅱ. 선행연구
Ⅲ. 연구 내용 및 방법
Ⅳ. 분석 사례 및 결과
Ⅶ. 논의 및 결론