文化関連語彙から見た機械翻訳の様相 ー『82年生まれ、キ厶・ジヨン』を中心にー

Abstract

This study analyzed machine translation errors and translation styles in translating Korean cultural vocabulary extracted from <Kim Jiyoung, Born 1982> into Japanese and Chinese. By comparing with human translation, it also considered the shortcomings and limitations of machine translation and proposed directions for improvement. The research found that in both Japanese and Chinese machine translations, the main issues were "errors due to meaning selection" and "lack of completeness." Mistranslations of polysemous words, homonyms, and vocabulary with unique Korean cultural backgrounds were particularly prominent. Additionally, there were errors caused by misinterpretation of the language's combinatory structure and situations where the unique linguistic habits of Korean lacked similar expressions in other languages, leading to information gaps. Regarding translationese, machine translation often relied too heavily on literal translation, resulting in the loss of subtle nuances and cultural meanings from the original text. Moreover, although transliteration was frequently used in Japanese translations, there were many cases where the cultural meanings of words were difficult to convey. To address these issues, the study concluded that machine translation systems need to be improved by considering linguistic ambiguity, cultural background, and contextual dependency. Adjusting algorithms to account for the multiple meanings of polysemous or homonymous words is crucial. Additionally, expanding databases of expressions related to specific cultures and systems, and adding supplementary information to explain cross-cultural elements, is necessary. It is also important to develop methods that translate naturally rather than relying on literal translations, and to strengthen algorithms that understand the appropriate vocabulary combinations based on context to ensure accurate translation.