基於檢索增強生成框架的聊天機器人搭配引導式對話與重複測驗以改善學生學習成效、參與度與動機;A RAG-based Chatbot with Guided Dialogue and Repeated Testing to Enhance Students′ Learning Performance, Engagement, and Motivation

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/98203

Please use this identifier to cite or link to this item: https://ir.lib.ncu.edu.tw/handle/987654321/98203

Title:	基於檢索增強生成框架的聊天機器人搭配引導式對話與重複測驗以改善學生學習成效、參與度與動機;A RAG-based Chatbot with Guided Dialogue and Repeated Testing to Enhance Students′ Learning Performance, Engagement, and Motivation
Authors:	林佳榆;Lin, Chia-Yu
Contributors:	資訊工程學系
Keywords:	RAG;重複測驗;語義相似度;學習活動;Retrieval-Augmented Generation;Repeated Testing;Semantic Similarity;Learning Activities
Date:	2025-07-01
Issue Date:	2025-10-17 12:29:00 (UTC+8)
Publisher:	國立中央大學
Abstract:	本研究開發 RevoTutor，一款基於檢索增強生成（Retrieval-Augmented Generation, RAG）架構的學習輔助聊天機器人（Chatbot），專門應用於機器學習與人工智慧課程設計。傳統 Chatbot 在專業領域常面臨回應準確性與上下文關聯性方面存在不足，影響學生的學習體驗與知識理解。RevoTutor 採用 RAG 技術，透過動態檢索機制獲取準確且相關的知識，提高回應的準確性與可靠性，並結合引導式對話和重複測驗機制，以提升學生的學習成效、參與度與動機。本研究說明RevoTutor的系統架構與功能設計，涵蓋課前預習、課後複習與即時疑問解答三大學習階段。課前預習透過引導式對話幫助學生建立清晰的知識框架，提升學習參與度與動機；課後複習運用重複測驗機制強化知識鞏固與轉移，促進長期記憶；即時疑問解答則透過 RAG 技術檢索進階學習資源提供高準確性的回應，支援學生在學習過程中的自主探索。實驗結果顯示，RevoTutor 在語義相似度優於現有多款商用與開源模型，而有引導式對話的學習模式加上重複測驗的複習機制能夠有效提升學生學習表現，並且提高學習成效。研究成果驗證了 RAG 技術結合教學策略在 AI輔助學習中的可行性與實用價值，展示生成式 AI 在提升學習品質、體驗和效果的潛力。;This study presents the development of RevoTutor, a learning-support chatbot based on the Retrieval-Augmented Generation (RAG) framework, designed for machine learning and artificial intelligence education. Traditional chatbots often face problems with response accuracy and context relevance in specialized subjects, which can affect students’ learning experience and understanding. RevoTutor uses RAG technology to dynamically retrieve accurate and related knowledge, improving the quality and reliability of responses. It also combines Guided Learning and Repeated Testing strategies to boost students’ learning performance, engagement, and motivation. The system supports three main learning stages: pre-class preparation, post-class review, and real-time question answering. Pre-class activities use guided dialogue to help students build a clear knowledge structure and stay engaged; post-class review uses repeated testing to reinforce memory and support knowledge transfer; and real-time queries use RAG to find advanced learning content and give accurate answers that help independent learning. Results show that RevoTutor outperforms several commercial and open-source models in semantic similarity, and its guided responses and testing method effectively improve learning results. These findings confirm the practical value of using RAG with teaching strategies in AI-supported learning and show the potential of generative AI to improve learning quality, experience, and outcomes.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	1	View/Open

社群 sharing

Loading...