일본 소설에 나타난 문화소의 한국어 기계번역 오류 유형 고찰―인간번역과의 비교 분석―
Analysis of Machine Translation Errors in Korean: Cultural Elements in Japanese Novels -Compared to Human Translation-
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
機械翻訳が多様な分野で幅広く活用され、翻訳家という職業も今後消えることになるという意見が多い。機械翻訳の場合、ニューラルネットワーク技術の発展によって正確性が高まり、人による翻訳と同等の水準に達したという指摘もあるが、文化的要素の翻訳には限界があるという指摘が多い。本稿では、日本の代表的な近現代小説に現れる文化的要素の韓国語機械翻訳と人による翻訳を語彙と慣用句を中心に比較分析を行った。その結果、機械翻訳が人による翻訳に比べて先行研究とは異なり、脱落エラーと音訳の翻訳に誤りが多かった。正確性、一貫性、可読性の面で機械翻訳が人による翻訳より正確度がはるかに低く、深刻なエラーも多かった。慣用句の場合、さらに深刻で意味が全く理解できないほどだった。また、現代小説に比べて近代小説の誤りが非常に多様で多くみられたが、今後近代語に関連するデータの蓄積が切実である。近年急速に機械翻訳の品質が良くなっているため、今後多様なデータが蓄積されれば機械翻訳の性能も向上し、文化的要素の機械翻訳の正確度も高くなると考えられる。
Abstract
As machine translation continues to be widely adopted across various fields, concerns persist regarding the potential obsolescence of human translators. While some argue that advancements in artificial neural network technology have enhanced the accuracy of machine translation to levels comparable to human translation, others contend that significant limitations remain, particularly in the translation of cultural nuances. This study examines Korean machine translation and human translation in representative modern and contemporary Japanese novels, comparing the results with findings from previous research. A high frequency of errors was identified, particularly in omission and mistranslation, affecting meaning and readability. Machine translation remains significantly less reliable than human translation regarding accuracy, consistency, and fluency, often producing severe errors. Translating idiomatic expressions is problematic, frequently resulting in nonsensical or incoherent output. Additionally, errors were more prevalent in contemporary novels than modern ones, highlighting the urgent need for extensive datasets covering contemporary linguistic usage. Although machine translation quality has improved rapidly, further advancements will depend on the availability of diverse and extensive datasets. Expanding such datasets is essential to enhancing machine translation performance, particularly in accurately conveying cultural and contextual elements.