Earticle

초록

영어

Researchers have reported positive effects of using pivot languages in improving the quality of statistical machine translation for western languages. The goal of this paper is to confirm whether the quality of English-to-Korean machine translation improves by using a pivot language, and to find out the best pivot language for English-to-Korean machine translation. For this, we carried out translation experiments using Google Translate; i.e., we tested the English-to-Korean machine translation of 83 sentences across 35 pivot languages. The test sentences were carefully controlled with reference to Levin's (1993) lexical semantic classification of English verbs. The log-linear regression analysis reveals that the quality of English-to-Korean machine translation improves when Japanese is used as a pivot language. This result is particularly interesting, since both English-Japanese and Korean-Japanese pairs have lower BLEU scores than the Korean-English pair. This suggests that the BLEU scoring system should be reevaluated and revised for future application to machine translation of natural languages.

목차

 1. 서론
 2. 이론적 배경
  2.1. 기계 번역 연구의 동향
  2.2. 중간 언어 방식
 3. 연구방법
  3.1. 효율성 평가
  3.2. 중간 언어 선정
  3.3. 영어 동사 분류
 4. 실험 결과: 교차표의 작성
 5. 로그선형 회귀분석
 6. 결론
 참고문헌
 부록

저자

• 박경리 [ Park, Kyoung-lee | 한국외국어대학교 ]
• 조수연 [ Jo, Sue-yeon | 한국외국어대학교 ]
• 전종섭 [ Jun, Jong-sup. | 한국외국어대학교 ]

참고문헌