Comparison of Korean-Russian punctuation of Machine Translation and Human Translation and its practical implications

다국어 초록 (Multilingual Abstract)

Recently Neural Machine Translation(NMT) has greatly improved machine translation and raised the debate about expectations and limitations on the machine translation quality. Google Neural Machine Translation (GNMT) was first enabled for 8 languages: to and from English and French, German, Spanish, Portuguese, Chinese, Japanese, Korean and Turkish in 2016. In March 2017, three additional languages were enabled: Russian, Hindi and Vietnamese along with Thai for which support was added later(Wikipedia). This paper focuses on inadequate punctuation in the sentence unit of machine translation. This paper compares the punctuation translation in the Google Neural Network Translation Service from Korean to Russian with the punctuation translation strategy of human translation to identify the problems and causes of MT and to suggest a practical implication. Subject of analysis is the Internet Newspaper, translated from Korea to Russian. Parallel corpus of Korean-Russian Human Translation(HT) and Machine Translation(MT) was analyzed by UAM Corpus tool for quantitative and qualitative analysis of punctuation translation.