An Empirical Study of Multi-layered Automatic Book Classification System in Library

• Main Authors: Hsin-Yu Hsieh, 謝心妤
• Other Authors: 郭俊桔
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/nd99nq

碩士 === 國立中興大學 === 圖書資訊學研究所 === 107 === The basic task of the library lies in the cultural collection and providing information. In order to achieve this goal, the library needs to carry out the basic work of preserving the books and materials, that is, the classification and cataloguing work. The current library organizes the books and materials, and is still classified by the cataloging staff. They spends a lot of time and spirit, and works on the books of classification and organization cataloging. Such manual classification work are faced with the rapid increase of books and materials in various subject areas, and the highly compressed manpower time limit, have been powerless. If we can tyr to use the information technology to assist in the processing of book classification, we hope to speed up the process of book classification and reduce the pressure on library classification work. In this study we collects a large number of Chinese e-books in the library, based on the actual library classification structure of the library is used as the experiment’s standard. We use the original bibliographic data such as the title, abstract, and catalogue of the book, after tokenizing process of the file, extracting the features of the file, and constructing the classification model. Then we can conduct classification experiments and discover the effectiveness of traditional single-layered machine classifiers for common machine learning. At the same time, we try to find out the title, abstract, catalog, and combination data set of the book, which one is the best combination of content for automatic classification of books? This study further explores to combine the advantages of multiple single classifiers with a multi-layered automatic book classifier architecture under the dual pressure of a large number of bibliographies and diverse categories, and finds the best classifier combination for multi-layered automatic book classification. The experimental results show that the classification precision of the multi-layered automatic book classification system in this study can reach 97.26%. Compared with the previous experimental research, the traditional single-layered classifier has only about 82% performance, showing better classification efficiency.