This study examines the contextual understanding and reliability of AI translations through Japanese–Korean and Korean–Japanese translations of newspaper articles. Because the terms “kindai-teki/geundae-jeok”, “gendai-teki/hyeondae-jeok” are used differently in Japanese and Korean, the study assesses whether AI translation can accurately grasp context and render these terms correctly. The results indicate that prompt instruction improves the AI’s contextual understanding, but translation accuracy remains inconsistent, limiting its reliability for professional translators. Although prompts increase accuracy in some cases, they reduce it in others, underscoring the importance of careful prompt design. Therefore, AI is useful for understanding the overall content or producing draft translations of informational texts such as newspaper articles. However, caution should be exercised when precise word choice and consistency are important. To make the most of an AI translation, translators need prompt design skills as well as the language proficiency and background knowledge necessary to guide and control the system effectively.
慎治 坂野 (Tue,) studied this question.