Earticle

초록

영어

This study explores gender neutrality in AI-based machine translation (MT), focusing on the translation of the Korean third-person singular gender-neutral expression “그 애” (that person) into English by three widely-used AI-based translation models: GPT-4o, Gemini 2.5 Flash, and DeepL. Using a parallel corpus extracted from the Korean novel Concerning My Daughter (Kim Hye-jin, 2017), the analysis compares AI-generated translations with human translations under conditions lacking contextual gender cues. Results indicate significant differences in gender marking strategies among models. GPT-4o and DeepL frequently added gender-specific pronouns (e.g., “she,” “he”) absent from the original text, while Gemini preserved gender neutrality but utilized overly formal and unnatural expressions (e.g., “that person”) for literary contexts. Additionally, human translators maintained neutrality through diverse strategies, including proper nouns. This research highlights critical ethical issues in AI-based MT regarding gender bias and suggests future research to enhance translation quality and gender neutrality in creative texts.

목차

1. 서론
2. 이론적 배경
2.1. 한국어와 영어의 성별 지시 표현 차이
2.2. 성별 지시 표현과 번역에 대한 연구
2.3. 기계번역에서의 젠더 편향(gender bias)
3. 연구 방법
3.1. 분석 텍스트
3.2. 분석 절차
3.3. 분석 결과 검정 절차
4. 연구 결과
4.1. 성별 중립적 지시 표현의 번역 양상
4.2. 성별 중립성 유지 양상
4.3. 번역 선택 양상
5. 결론 및 시사점
참고문헌

저자

• 임진 [ Yim, Jin | 이화여자대학교 ]

참고문헌