E-book Reading Behavior Analysis and Recommendation Based on Hidden Markov Models

[[abstract]]資訊與網路科技的發展使得線上學習日益普及，近年來電子書技術與應用更發展蓬勃，相較於傳統紙本書籍，能提供學習者更豐富的多媒體資訊。在缺乏面授教學的線上學習環境中，適性化學習的重要性更加提升。適性化學習可以依據學生的個別化特性調整教授方式以增加學生學習成效。然而目前的研究卻鮮少提供可適性化的電子書閱讀系統的相關探討與應用。 本研究建立一可適性化的電子書閱讀系統，能針對不同學習者提供適性化的閱讀建議。根據學生閱讀行為的分析結果，我們根據不同的行為類別提供不同的閱讀建議，以適合不同學生的需要。本論文所提學生閱讀行為分析之演算法是基於隱式馬可夫鏈，透過擷取學生的動態閱讀行為，我們能針對學生章節與媒體的閱讀順序進行閱讀行為的分類。在推薦閱讀行為的步驟中，我們根據學生閱讀行為所探勘出的閱讀認真程度給予閱讀推薦。 實驗結果顯示本研究所提之學生閱讀行為分析與推薦演算法是可行的。在基於隱式馬可夫鏈的學生閱讀行為分類上，我們可以達到正確的分類結果；在閱讀行為推薦上，研究所提方法亦可針對不同學生作有效的推薦。因此，本研究的確可為未來適性化電子書閱讀系統提供一可行的研究方向，增加未來的線上閱讀環境的學習者閱讀行為分析與推薦。[[abstract]]With the rapid development of information and network technology, more and more teachers utilize e-books in the class which provide not only multimedia contents but also more interactive and friendly user interfaces and thereby change students’ reading behaviors. Therefore, exploring the design elements and principles of e-books and evaluating their corresponding effectiveness for teaching purposes are significant issues in the field of education. However, few existing works analyzed learners’ e-book reading behaviors or considered the adaptability of the e-book for the purpose of giving different functions and recommendations for different users. In this study, we try to build an adaptive e-book system which providing recommendations for e-book readers according to the analyzing results of their reading behaviors. The proposed algorithms of reading behavior analysis and recommendation are based on Hidden Markov Models, with which students’ dynamic reading behaviors can be understood by extracting the features of sequential reading actions and producing corresponding suggestions according to the results of reading behavior classification. By conducting both technical and subjective experiments, the proposed reading behavior analysis and recommendation system is proved to be feasible for students’ e-book reading.[[note]]碩