This study examines a method for structuralizing Korean oral folktale materials contained in the digital archive of The Great Collection of Korean Oral Literature so that they can be more stably identified, compared, and rearranged in an environment where generative AI is used. To this end, this study focuses on the tale type The Brother and Sister Who Became the Sun and the Moon. First, relevant materials were searched using AI, and the researcher then reviewed the results and confirmed forty-five texts as the object of analysis. Based on these forty-five texts, the preliminary structuralization results suggested by AI were examined. Items involving interpretive naming or inconsistent levels of classification were adjusted, and criteria for establishing event units were developed around observable actions and situations.
On this basis, the narrative was organized into seventeen event units: the mother’s departure, the encounter with the tiger, the tiger’s demand, the mother’s death, the tiger’s attempted intrusion, the children’s verification of whether the visitor is their mother, the tiger’s intrusion, the children’s escape, the children’s hiding, the tiger’s pursuit, the children’s response or deception, the children’s prayer, the children’s ascent, the transformation into the sun and the moon, the tiger’s attempted ascent, the tiger’s fall, and the origin of the red sorghum stalks. These criteria were applied to all forty-five texts. Events that appeared repeatedly were marked in the same way, while variant points in each text were recorded separately. The results show that event-unit-based structuralization enables materials with different titles to be identified again on the basis of their shared event sequences. It also makes it possible to compare variants as differences in the way specific event units are realized and to rearrange the materials according to various analytical categories. However, because AI-generated structuralization may include omissions or other errors, comparison with the original texts and correction by the researcher are essential.
This study is significant in that it presents a methodological possibility for transforming Korean oral folktale materials into research data that can be compared and accumulated through the use of generative AI. As a data-processing method that combines AI-based preliminary processing with human review and reconfiguration, this approach suggests that large-scale oral folktale collections such as The Great Collection of Korean Oral Literature can be expanded beyond a simple archive of original texts into a knowledge base for comparing event structures and patterns of variation by tale type. Furthermore, the ‘event unit schema’ and human-AI collaboration pipeline examined in this study demonstrate potential for systematic expansion into other voluminous tale types and international comparative literature.