LLM 기반 텍스트마이닝: 생성형 AI 활용 교육에 관한 연구를 중심으로

초록

This study collected and pre‑processed the titles and abstracts of 1,436 domestic research papers on generative AI in education published since 2023, with the aim of systematically analyzing emerging trends in its educational applications. Using GPT‑4o for keyword extraction, we performed K‑Means clustering (n = 5) and latent Dirichlet allocation (LDA) topic modeling (five topics). The integrated analysis of LDA topics and clusters converged on five thematic axes—arts convergence and creativity, language education and prompt strategies, elementary feedback and learning effectiveness, interactive learning‑environment design, and curriculum ethics and governance—indicating that generative AI research is expanding from language‑ and arts‑centered subjects into STEM fields while simultaneously deepening pedagogical strategies and ethical‑policy discourse. Methodologically, the LLM‑based approach captured contextual meanings and implicit conceptual links that conventional TF‑IDF–LDA methods often overlook, thereby enhancing the coherence of topic interpretation. By quantitatively mapping megatrends in generative‑AI education research and empirically assessing both the strengths and limitations of LLM‑driven text analysis, this study provides foundational evidence for future policy design and follow‑up research.

키워드