Semantic Similarity Analysis of Narrative Scenarios Generated by Generative AI: A Convergence-Oriented Study of Narrative Structures and Creative Implications for Animation Scenario Research Using Text Mining Techniques

This study quantitatively analyzes scene-level semantic similarity in five short animation scenarios generated by generative artificial intelligence (ChatGPT-4.0). The primary aim is to verify the structural differentiation between scenarios and their internal cohesion, while also interpreting these features from an integrated perspective. Using TF-IDF and cosine similarity, an inter-scenario distance matrix was constructed, revealing that Scenario ID 3 formed an independently differentiated narrative structure, whereas IDs 1, 2, 4, and 5 clustered within a narrower semantic range.<br/> The analysis revealed distinctive characteristics across scenarios and cohesion within each scenario. Some showed independent structural patterns, while most followed traditional narrative forms, indicating that AI-generated narratives possess a certain degree of structural completeness rather than arbitrary outcomes. Unlike previous research, which has focused primarily on qualitative evaluations, this study offers empirical verification of narrative structure through quantitative analysis.<br/> Furthermore, this research combines scientific text-mining techniques with the artistic context of animation scenarios, suggesting new possibilities for interdisciplinary studies between engineering and the arts. Despite some limitations, this study demonstrates the structural patterns of AI narratives and highlights their creative potential for academic and practical applications. Future research will require a more integrated approach combining large-scale data and advanced methods.