基于知识组织的图书知识服务系统的设计与实现

【作者】 李戈；

【导师】 吴江琴； 鲁伟明；

【作者基本信息】 浙江大学 ， 计算机科学与技术， 2015， 硕士

【摘要】 CADAL数字图书馆已扫描数字化图书260多万册,为人们获取知识提供了丰富的来源。然而,图书虽然被数字化了,可以方便在电脑前阅读,但其仍然采用了传统纸质图书的阅读方式,也就是按图书目录来阅读每本图书。这并没有极大地发挥数字图书可分解、可快速定位的优点。鉴于此,我们希望能将数字图书中的知识单元抽取出来,并将这些知识单元有机组织,使用户更容易获取知识。本文从目录的结构化学习角度出发,提出了一种将结构化碎片整合为复杂层次结构的方法,并通过时序关系分析与知识单元抽取对层次结构进行拓展,最后将其应用于图书知识服务系统中。本文的主要工作包括1)研究了结构化学习以及知识单元抽取等相关技术的基本方法。2)基于近邻传播模型,提出了一种从大量结构化碎片中学习复杂层次结构的方法,通过实验验证了该方法在多个维度都优于基本的近邻传播模型,最后将该方法应用于从大量图书目录构建知识层次网络中。3)通过对于知识单元的提取以及用户学习拓扑结构的分析,对于原有知识层次网络进行了拓展。4)将学习得到的层次模型应用于图书知识服务系统中,并完成了知识服务系统的知识库构建,数据存储,基础服务,知识服务以及用户管理等功能。5)最后实现并测试了整个系统,并将其上线运行。更多还原

【Abstract】 There are lots of knowledge in the traditional book, but there is no way to search them quickly and accurately. So we need a solution to fragment the traditional book and re-construction them. More than 2.6 million books in digital library CADAL provide us the data source. And that we extract the rich knowledge from the books will enhance our reading and knowledge service.So, we are hoping to extract the knowledge units in digital books, and reconstruct knowledge units to ease user. In this paper, we propose a structural learning method to build a big and complex hierarchical structure from fragments. And we extend this structure though topological analysis and knowledge unit. At last, we apply this complex structure to book knowledge service.The paper regard the actual project as the background. We develop and test the China Knowledge Centre for Engineering Sciences and Books Knowledge service. The main work includes 1) Researching the related work about the structural learning and annotation.2) Based on affinity propagation, we propose a structural learning method and we construct the hierarchical structure from the catalog in books based on this method.3) We proposed a method based on Word2Vec+AP_Cluster+TextRank to label the picture and mining the topological structure on the catalog of the books. Then, we extend the hierarchical structure with picture and topological structure.4) Applying hierarchical structure to our service.5) Giving the implementation and test of the entire system. The system is available on the internet.更多还原