한국소설 영어번역서에 부여된 주제명의 현황 분석과 자동분류에 관한 연구

초록

This study analyzes the subject headings of 492 English translations of Korean fictions and evaluates machine learning-based automatic classification models. Bibliographic data were collected from the Digital Library of Korean Literature and WorldCat. Subject heading frequencies and FAST facet distributions were visualized, and key labels were selected for multi-label classification. Among various models, deep learning models using summaries as features showed the highest performance (F1 = 0.62, AUC = 0.89), with AUC values above 0.8 for 9 out of 10 labels. Additionally, based on ROC curves and confusion matrices, the study identified labels with lower performance and explored the relationships between certain labels. This study demonstrates the potential of deep learning models for classifying subjects in translated Korean literature.

키워드