´ëÇѾð¾îÇÐȸThe Linguistic Association of Korea

ÇÐȸÁö

  • Ȩ
  • ÇÐȸÁö
  • ³í¹®ÀÚ·á½Ç

³í¹®ÀÚ·á½Ç

Á¦¸ñ A Comparative Error Analysis of Neural Machine Translation Output: Based on Film Corpus
ÀúÀÚ Sungran Koh
±Ç/È£ Á¦30±Ç / 1È£
Ãâó 157-177
³í¹®°ÔÀçÀÏ 2022-03-31
ÃÊ·Ï Koh, Sungran. (2022). A comparative error analysis of neural machine translation output: Based on film corpus. The Linguistic Association of Korea Journal, 30(1), 157-177. This study aims to analyze translation from Korean to English in three mainstream machine translation (MT) systems in Korea and to classify the major error problems of the MT systems. To do this, first, the Korean script of the film Minari (2021) was collected and translated by three major machine translators (Google Translate, Papago, and Kakao i). Then, the translation output of the three mainstream online translation systems was manually evaluated by humans. Next, MT errors in Korean to English were classified into four categories: missing words, word order, incorrect words, and unknown words. The incorrect words were subcategorized into sense, incorrect form, and extra words. The most frequent type of incorrect word error was incorrect disambiguation (subject) and wrong lexical choice in terms of sense. Based on these findings, some suggestions are to use more developed machine translation for both MT system developers and Korean English as a Foreign Language(EFL) learners. This study sheds light on the quality of current MT systems based on the error analysis of this data and offers EFL learners insights into using MT systems better.
ÆÄÀÏ PDFº¸±â  ´Ù¿î·Îµå