한노 기계번역의 오류 유형화 및 품질 개선을 위한 프리에디팅(pre-editing) 규칙 제안 (A Study on Error's Type and Pre-Editing Rules to Enhance Output Quality of Korean-Russian Machine Translation)

서지정보

· 발행기관 : 한국외국어대학교 통번역연구소
· 수록지 정보 : 통번역학연구 / 23권 / 3호 / 291 ~ 327페이지
· 저자명 : 한현희

초록

This paper aims to categorize the errors of the Korean-Russian machine translation and establish pre-editing rules to improve the quality of machine translation. Result shows that types of errors include vocabulary and context error, typo, others, verbal error, (vocabulary/phrase) omission, (vocabulary/phrase) structural error and an error on a proper noun and number/time/date error in following order. Based on these, 10 pre-editing rules have been established and suggested. 1) The formal name must be used instead of an abbreviation for a particular organization. 2) numerical writing methods which use Korean units must be avoided. 3) address must be divided for each unit by using a comma. 4) in the case of date, correct information must be provided to identify the context such as year, month, date and etc.
5) the usage of polysemy or homonym must be avoided. 6) the meaning must be clarified. 7) the usage of dependent nouns which can cause confusion in identifying the context must be avoided 8) addition of unnecessary information(symbol) must be avoided 9) abbreviated subject must be clarified. 10) the sentence structure must be simplified. As a result of using these rules, it has been confirmed that the errors of machine translation have been significantly improved.

영어초록

This paper aims to categorize the errors of the Korean-Russian machine translation and establish pre-editing rules to improve the quality of machine translation. Result shows that types of errors include vocabulary and context error, typo, others, verbal error, (vocabulary/phrase) omission, (vocabulary/phrase) structural error and an error on a proper noun and number/time/date error in following order. Based on these, 10 pre-editing rules have been established and suggested. 1) The formal name must be used instead of an abbreviation for a particular organization. 2) numerical writing methods which use Korean units must be avoided. 3) address must be divided for each unit by using a comma. 4) in the case of date, correct information must be provided to identify the context such as year, month, date and etc.
5) the usage of polysemy or homonym must be avoided. 6) the meaning must be clarified. 7) the usage of dependent nouns which can cause confusion in identifying the context must be avoided 8) addition of unnecessary information(symbol) must be avoided 9) abbreviated subject must be clarified. 10) the sentence structure must be simplified. As a result of using these rules, it has been confirmed that the errors of machine translation have been significantly improved.

참고자료

· 없음