아랍어-한국어 기계번역에 관한 고찰

This study aims to analyze differences between Google Neural Machine Translation(GNMT) and Human translation(HT) in Korean translation of Arabic non-literary text. GNMT was performed on Arabic non-literary two texts consisting of 789 words, and differences between GNMT and HT were analyzed on lexical, syntactic, pragmatics, cultural-informational aspects. Details include spacing, punctuation, choice of Korean vocabulary appropriate for the context of when Arabic homograph is us, completeness of sentences, match of subject and predicate, translation of pronouns and demonstrative pronouns, proper sentence arrangement, conversion of part of speech in translated text, appropriate translation of Arabic polysemy into Korean for the context, use information in translation, natural Korean, Korean readability. In addition, post-editing and pre-editing were checked by illustrative cases to see if they could contribute to improving the quality of GNMT. Further research is needed on whether GNMT can be used for educational and commercial translation purposes.