建構圖書館個人化推薦系統-以國立政治大學為例

近年來由於資訊的快速增長，圖書館普遍面臨圖書借閱率下降的問題，而讀者如何從不斷擴增書籍的圖書館中找到符合自己喜歡的書籍也是一大考驗。因此，本研究應用政大新啟用之圖書館服務平台Ａlma所記錄之讀者圖書借閱記錄，並以圖書資訊學資訊組織角度，將個人及整體的圖書借閱紀錄應用於發展以內容為基礎及以協同過濾為基礎的個人化圖書推薦系統，利用主動且個人化的方式為讀者提供圖書推薦服務。本研究的推薦結果分為兩大功能呈現：「當月新書推薦」、「您可能感興趣推薦（舊書推薦）」，並予以「借閱記錄視覺化」輔助讀者了解個人借書行為習慣。研究結果顯示，本研究所發展的圖書館個人化書籍推薦系統功能中，以內容為基礎的推薦書籍多為符合個人閱讀偏好興趣的圖書，能加深讀者對於感興趣領域圖書的見聞;以協同過濾推薦之書籍則以多元化的閱讀類型推薦，可擴展讀者的閱讀視野。此外，本研究所發展的圖書館個人化書籍推薦系統可有效提升讀者前往圖書館借閱書籍的動力及增進其對於圖書館新書資訊的掌握，對於讀者及圖書館皆呈現正向的影響。再則個人借閱記錄視覺化的部分則可有效幫助讀者了解自己的閱讀行為及喜好脈絡。未來，個人化書籍推薦系統若能加入借閱時間作為推薦考量，將能使系統更加完善，更能滿足讀者當下的閱讀喜好。

關鍵字

推薦系統；數位圖書館；以內容為基礎演算法；以協同過濾為基礎演算法；矩陣分解法；資訊視覺化

並列摘要

Due to the rapid growth of information in recent years, libraries commonly face the problem of declining book borrowing rates, as readers find it difficult to find the book they need from an ever-increasing number of books. Hence, this study applied the book loan data from the library service platform Alma, which was newly launched by National Chengchi University Library, to record the borrowing records of readers. Taking the perspectives in Cataloging and Classification of Library and Information Science, this research was conducted by applying library loan records to two kinds of algorithm, based on content-based filtering (CBF) and collaborative filtering (CF). The results were shown via the functions “The Month New Arrival Books” and “You Maybe Like” on the interface. In addition, the system included “Visualized Borrowing Records” to help readers know their borrowing habits. The experimental results show that within the system’s functions, content-based filtering (CBF) recommended books were able to meet individual reading preferences, deepen readers’ insights into the topics of interest, while collaborative filtering (CF) recommended books across a wider spectrum, expanding readers’ horizons. Moreover, the result of this research confirmed that the library personalized recommender system effectively encouraged readers’ motivation to borrow books from the library, while enhancing readers’ understanding of the new books. This generated a positive impact on both the readers and the library. Furthermore, the visual part of personal borrowing records can help readers understand their reading behavior and preferences effectively. In the future, if the personalized book recommendation system can factor the length of time of each loan into consideration, it will further complete the system and meet the readers’ preferences.