AI在教育研究領域的應用系列（十九）：NotebookLM串接Gemini進行深度分析

Applications of AI in Educational Research Series (XIX): Integrating NotebookLM With Gemini for In-Depth Analysis

本研究旨在探討NotebookLM串接Gemini進行深度分析之研究流程與其在系統性文獻回顧與研究分析中的應用模式。隨著生成式人工智慧技術逐步導入學術研究與文獻分析工作，人工智慧工具在文獻整理、主題分析與研究架構建構方面展現高度潛力。本文以實作導向方式，說明NotebookLM建立文獻資料庫、文獻主題整理、研究趨勢分析、研究缺口分析與文獻回顧架構建構之流程，並進一步說明NotebookLM串接Gemini進行深度分析、研究架構推論與研究假設發展之操作流程與應用方式。實作結果顯示，NotebookLM在文獻摘要、主題整理與資料來源溯源方面具有良好效率，能協助研究者快速掌握文獻內容與研究趨勢；而Gemini在跨文獻整合分析、概念比較與研究輸出方面具有較強之推論與內容生成能力。兩者整合應用可有效提升文獻回顧效率與研究發展效率。然而，生成式人工智慧工具在文獻分析過程中仍可能存在資料解讀偏誤與研究判讀問題而需研究者專業判斷等限制，因此較適合採用人機協作研究模式，由人工智慧協助資料整理與初步分析，而研究者負責理論詮釋與研究品質控管，以兼顧研究效率與研究品質。

This study explores the research workflow of integrating NotebookLM with Gemini for in-depth analysis and examines its application in systematic literature review and research analysis. With the increasing integration of generative artificial intelligence into academic research and literature analysis, AI tools have demonstrated strong potential in literature organization, thematic analysis, and research framework development. This study adopts a practice-oriented approach to illustrate the workflow of using NotebookLM to build a literature database, organize research themes, analyze research trends, identify research gaps, and construct a literature review framework. Furthermore, it demonstrates how NotebookLM can be integrated with Gemini to conduct deeper analysis, research framework development, and hypothesis generation. The results indicate that NotebookLM performs efficiently in literature summarization, theme organization, and source grounding, helping researchers quickly understand research trends and literature content. Gemini, on the other hand, shows stronger capabilities in cross-document synthesis, conceptual comparison, and research output generation. The integration of the two tools can significantly improve the efficiency of literature review and research development. However, generative AI tools may still produce misinterpretations or require human judgment in research interpretation. Therefore, a human AI collaborative research model is recommended, in which AI assists with data organization and preliminary analysis, while researchers remain responsible for theoretical interpretation and research quality control.