探討基於人工智慧的概念圖結合QAR機制促進批判性EFL寫作;Investigation of AI-based Concept Map with Question Answer Response (QAR) to Facilitate Critical Level EFL Writing

NCU Institutional Repository > 資訊電機學院 > 網路學習科技研究所 > 博碩士論文 > Item 987654321/98362

請使用永久網址來引用或連結此文件: https://ir.lib.ncu.edu.tw/handle/987654321/98362

題名:	探討基於人工智慧的概念圖結合QAR機制促進批判性EFL寫作;Investigation of AI-based Concept Map with Question Answer Response (QAR) to Facilitate Critical Level EFL Writing
作者:	周一輝;Pradana, Ariel Adi
貢獻者:	網路學習科技研究所
關鍵詞:	概念圖;問答;EFL寫作;人工智慧;Concept Map;Question Answering;EFL Writing;Artificial Intelligence
日期:	2025-06-23
上傳時間:	2025-10-17 12:40:56 (UTC+8)
出版者:	國立中央大學
摘要:	對於以英語為外語(English as a Foreign Language, EFL)的高中學生來說，具備歐洲共同語言參考標準(Common European Framework of Reference for Languages, CEFR)B2至C1等級的英語寫作能非常重要。「批判性EFL寫作」要求文章具有嚴謹的架構，並且需要在多個層面展現出熟練的能力，包括任務達成度、連貫性、詞彙變換能力以及文法精確度。然而，許多現有的EFL寫作工具主要著重於語法和詞彙的輔助，但往往無法同時兼顧內容發展與寫作連貫性，因此不足以有效支援批判性EFL寫作的教學。本研究開發了一個先進的系統，整合了 GPT-4o產生的概念圖和問答回應機制(Question Answer Response, QAR)，以加強批判性EFL寫作教學。進行了一項為期六週的準實驗，對象為197名印尼高中生，分為實驗組與控制組。實驗組的學生使用根據學生與GPT-4o的寫作所自動生成的概念圖，並運用QAR來比較概念圖並組織寫作。隨後，實驗組接受基於概念圖的個別化鷹架來修改寫作，而控制組則採用以教師為中心的鷹架。結果顯示，實驗組在學習成就與實作表現上均優於控制組，特別是因為實驗組加入了個人化鷹架組織寫作，使得在寫作連貫性方面表現特別突出。此外，在修改寫作過程中加入了GPT-4o生成的節點、問題與回應，對於任務達成度具有正向的影響。實驗組在實作過程中也展現出顯著較高的字數與更豐富的辭彙變換能力，這主要歸因於結合了GPT-4o生成的節點與學生的回答。此外，問卷調查的結果顯示實驗組對學習系統持正面看法。這些發現突顯了概念圖、QAR和個別化鷹架在提升批判性EFL寫作教學中的成效。未來的研究應考慮整合語法回饋、擴展概念圖生成的資料集，並整合協作寫作的機制。;Achieving B2 to C1 of English writing proficiency on the Common European Framework of Reference for Languages (CEFR), is critical for senior high school English as a Foreign Language (EFL) students. This ′critical level EFL writing′ demands well-structured text, demonstrating proficiency across multiple dimensions, including task achievement, coherence, lexical resources, and grammatical accuracy. However, many existing EFL writing tools primarily focus on grammar and vocabulary assistance but often fall short in addressing both content development and writing coherence, thus inadequately supporting critical level EFL writing instruction. This study developed an advanced system integrating GPT-4o-generated concept maps and Question Answer Response (QAR) to enhance critical level EFL writing instruction. A six-week quasi-experiment was conducted, involving 197 Indonesian senior high school students, divided into Experimental Group (EG) and Control Group (CG). EG students used automatic generated concept maps based on both students and GPT-4o’s writing, and QAR for concept maps comparison and writing organization. Subsequently, EG received concept map-based personalized scaffolding to revise their writing, while CG used teacher-centered scaffolding. The results revealed that EG outperformed CG in both learning achievement and practice performance, particularly in writing coherence, which attributable to personalized scaffolding. Furthermore, incorporating GPT-4o-generated nodes, questions, and responses during writing revision positively affect task achievement. EG also exhibited significantly higher word count and richer lexical resources during practice, primarily driven by incorporating GPT-4o-generated nodes and students’ answer. Moreover, the questionnaire revealed EG’s positive perception of the learning system. These findings highlight the effectiveness of concept maps, QAR, and personalized scaffolding to improve critical level EFL writing instruction. Future studies should consider integrating grammar feedback, expanding the dataset for concept map generation, and integrating collaborative writing setup.
顯示於類別:	[網路學習科技研究所 ] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	15	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....