1. 서론

2. 이론적 배경

3. 연구 방법

4. 연구 결과 및 논의

5. 결론

This study constructed a corpus of 300 Korean utterances from Lee Isaac Chung's film Minari (2020) and examined its diasporic emotional structure by character and generation, using a deep learning model (KcELECTRA-based KOTE) in tandem with a lexicon-based analysis using a Korean emotion word dictionary of 24 categories. The findings revealed three key structures. First, characters exhibited different emotional patterns during the same conflicts: Jacob's anger, Monica's disappointment and anxiety, Soonja's caring, and the children's embarrassment and unease. Second, the level of emotional expression in Korean decreased gradually from the grandparent to the child generation, as evidenced by two independently derived measures. Third, while the lexicon-based analysis revealed no explicit emotion-related words in more than 80% of the utterances, the deep learning model predicted distinct emotions for the majority of these lexically silent utterances, indicating that emotion in Minari is conveyed through tone and context rather than through explicit vocabulary. This study, framed as computer-assisted close reading, proposes a working model that combines computational analysis and humanistic interpretation.