基於間隔效應與知識追蹤之適性化學習演算法系統設計與應用：以多益英語學習為例;Design and Application of an Adaptive Learning Algorithm System based on Spacing Effects and Knowledge Tracing: Case Study on TOEIC English Learning

NCU Institutional Repository > 資訊電機學院 > 網路學習科技研究所 > 博碩士論文 > Item 987654321/93414

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/93414

題名:	基於間隔效應與知識追蹤之適性化學習演算法系統設計與應用：以多益英語學習為例;Design and Application of an Adaptive Learning Algorithm System based on Spacing Effects and Knowledge Tracing: Case Study on TOEIC English Learning
作者:	徐世凡;Hsu, Shih-Fan
貢獻者:	網路學習科技研究所
關鍵詞:	學習分析;教育資料探勘;知識追蹤;第二外語學習;Learning Analytics;Educational Data Mining;Knowledge Tracking;Second Language Acquisition
日期:	2023-07-26
上傳時間:	2024-09-19 16:57:53 (UTC+8)
出版者:	國立中央大學
摘要:	英語作為全球通用的語言之一，在現今全球化的時代學習英語技能變得極為重要，而精通第二語言需要時間並且持續地學習和積極練習。適應性間隔複習方法對於第二語言學習具有正向影響，可提高學習者的長期記憶和學習品質。而透過知識追蹤演算法可以針對學習者的學習軌跡進行知識建模，了解學習者當下的知識掌握程度，協助學習者針對弱點進行練習與複習。本研究拓展了以往的知識追蹤演算法，開發了Tagorithm知識追蹤演算法。並於兩所臺灣北部某國立大學招募學習者進行實驗，研究對象共計61人，對照組29人、實驗組32人。對照組練習時，系統從題庫中隨機抽取題目練習與複習；實驗組練習時，系統依照Tagorithm知識追蹤預測的知識點掌握程度推送學習者需要練習和複習的題目。本研究實驗進行12周，於第一周與第十二周分別進行前測與後測，包含英語能力測驗、線上自我調節學習問卷以及自我導向學習問卷。在實驗後，使用英語能力前測與練習紀錄作為模型的訓練集，英語能力後測作為測試集以評估模型的預測能力。研究結果發現，Tagoritm知識追蹤演算法能夠準確預測學習者們各知識點的掌握程度，使用預測結果推薦題目的實驗組在學習成效有顯著的提升以及反應時間有顯著的下降。整體而言，相較於對照組的隨機出題方式，Tagoritm知識追蹤演算法能夠幫助學習者們有效率的提升學習成效。並提供視覺化儀表板呈現各知識點的掌握程度，讓學習者與管理者了解學習狀況。而經過實驗後對於線上自我調節學習能力以及自我導向學習能力兩組皆無顯著差異。最後，依據訪談結果，實驗參與者對於專業性與實用性面向表示肯定，並認為適性化間隔複習機制能夠幫助他們有效地複習並提升英語能力。 ;English, as one of the globally recognized languages, has become increasingly important in the era of globalization. Achieving fluency in a second language requires time, continuous learning, and active practice. Adaptive spaced repetition methods have positively affected second language learning, enhancing learners′ long-term memory and learning quality. Through knowledge tracing algorithms, it is possible to model learners′ knowledge trajectories, understand their current grasp of knowledge, and assist them in practicing and reviewing areas of weakness. This study expands on previous knowledge tracing algorithms and develops the Tagorithm knowledge tracing algorithm. It recruited 61 participants from two universities in northern Taiwan, with 29 in the control group and 32 in the experimental group. During practice sessions, the control group randomly selected questions from a question bank for practice and review, while the experimental group received practice and review recommendations based on Tagorithm′s predicted knowledge mastery. The experiment lasted 12 weeks, with pre-tests and post-tests conducted in the first and twelfth weeks, including English proficiency tests, online self-regulated learning questionnaires, and self-directed learning questionnaires. After the experiment, the English proficiency pre-tests and practice records were used as the training set, and the English proficiency post-tests were used as the test set to evaluate our proposed model′s predictive ability. The research findings indicate that the Tagorithm knowledge tracing algorithm accurately predicts learners′ mastery of various knowledge points. The experimental group, which received practice recommendations based on the predictive results, showed significant improvement in learning outcomes and significantly reduced response times. Overall, compared to the control group′s random question approach, the Tagorithm knowledge tracing algorithm efficiently improved learning outcomes for learners and provided a visualized dashboard displaying the mastery levels of various knowledge points, enabling learners and administrators to understand learning progress. Following the experiment, there were no significant differences in online self-regulated learning and self-directed learning abilities between the two groups. Finally, based on the interview results, participants appreciated the algorithm′s professionalism and acknowledged the benefits of adaptive spaced repetition mechanism in helping them efficiently review and improve their English proficiency.
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	46	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.....