「協作學習」強調學生對認知、情意與後設認知資源的協調運用，而知識翻新主張學生需共同承擔推進社群知識的責任，透過協作創造具價值的知識。在知識翻新教學中，對話分析是理解想法發展歷程的重要方法，然而傳統質性分析方式效率低且難以規模化，限制了其在實務現場的應用潛力。為回應此挑戰，本研究開發一套整合自然語言處理技術的「知識翻新學習歷程分析與自動回饋鷹架模組」，結合知識翻新理論、對話分類與序列分析方法，即時分析學生在學習討論中的想法發展，並透過儀表板與即時建議協助師生理解與優化學習歷程。模組設計採跨領域合作，系統包含前後端架構、分類模型與互動介面，並以國小自然科探究課程為應用場域進行初步測試。 本研究以新北市某國小高年級學生為對象，透過問卷與開放式回饋，分析學生對模組的科技接受度與使用經驗。結果顯示，學生普遍認為模組具備高度知覺有用性與易用性，願意持續使用於討論學習中，開放式回饋亦指出模組有助於提升討論理解、自我反思能力與參與意願。研究結果支持本模組作為提升知識翻新教學實施效能的可行方案，未來將持續優化系統並拓展不同學科與情境之應用。 ;Constructive collaborative learning emphasizes students’ coordinated use of cognitive, affective, and metacognitive resources, while knowledge building advocates for shared responsibility in advancing community knowledge through collaborative creation of valuable ideas. In knowledge-building pedagogy, dialogue analysis is a crucial method for understanding learning processes. However, traditional qualitative analysis approaches are time-consuming and difficult to scale, limiting their practical application. To address this challenge, this study developed a "Learning Dialogue Analysis Module" that integrates natural language processing （NLP） technologies. Grounded in knowledge-building theory, this module combines dialogue classification and sequential analysis methods to provide real-time analysis of students’ performance in collaborative discussions. Through a dashboard and instant suggestions, it supports teachers and students in understanding and optimizing learning processes. The module was co-designed through interdisciplinary collaboration and includes a front-end/back-end system, classification models, and interactive interfaces. It was pilot-tested in an elementary school science inquiry course. This study targeted upper-grade students at an elementary school in New Taipei City, using questionnaires and open-ended feedback to assess students’ technology acceptance and user experience. Results indicate that students generally perceived the module as highly useful and easy to use, and expressed willingness to continue using it in learning discussions. Open-ended responses also highlighted its benefits in enhancing discussion comprehension, selfreflection, and participation. The findings support the module as a viable solution to improve the implementation of knowledge-building pedagogy. Future work will focus on optimizing the system and extending its application to various subjects and learning contexts.
