국문장편소설의 감정에 대한 정성적·정량적 연구 시론 -<창선감의록>과 <소현성록>을 중심으로- = Qualitative and quantitative research on emotions in Korean full-length novels - Focusing on Changseongamuirok and Sohyeonseongrok -

다국어 초록 (Multilingual Abstract)

This study aims to establish a new methodology for studying emotions in Korean novels by combining qualitative research in humanities and quantitative research in digital humanities. The process of this study consisted of three stages: data collection and construction, deep learning emotion prediction model creation and emotion prediction, and comparison and verification of emotion prediction results. Through this, the ‘KOTE model’(44 categories), ‘Kang woo-kyu model(v.1.1.0.)’(18 categories), and ‘Kang Woo-kyu model(v.1.1.1.)’(18 categories) were constructed. In addition, the emotion prediction results of the three models for Changseongamuirok and Sohyeonseongrok were compared to verify their performance. As a result, the ‘Kang woo-kyu model(v.1.1.1.)’ showed the highest accuracy. This is because the ‘Kang woo-kyu model(v.1.1.1.)’ learned data in which the researcher qualitatively identified the emotions of Changseongamuirok. Therefore, the use of the ‘Kang woo-kyu model(v.1.1.1.)’ can be a way to save time and cost for researchers by predicting the emotions of Korean full-length novels relatively accurately. In addition, by setting different emotion classification systems for the same subject according to the purpose, needs, and thoughts of researchers or research groups and providing a starting point for creating each deep learning emotion prediction model, I believe that this will be an opportunity to merge the multi-layered and diverse thoughts and achievements of literary research with digital (deep learning).