【作者】 衣英楠；

【导师】 马军；

【作者基本信息】 山东大学 ， 计算机系统结构， 2005， 硕士

【摘要】 知识管理是近年来国际学术界和产业界研究的热点问题,它在实践中尤其是商业企业中得到了日益广泛的应用,创造出了巨大的商业价值。从计算机学科的角度看待知识管理,它是以互联网和信息技术为基础,帮助企业或组织对机构相关的知识资源进行明晰化、系统化的管理,定位组织内拥有专门技能的人,建立团队协作的专家网络,使组织内部的人们快速而方便的访问和学习到所需要的信息和知识,以实现最佳的决策,运用集体的智慧来提高整体的协作和创新能力。 当前,对知识管理的研究尚处于探索之中,目前所开发的知识管理系统存有很多不足。以知识管理的一个方面——数字文档管理系统(DDMS)中的知识检索为例:所谓知识检索,就是用户为了求解某个问题查询多个领域的相关文档,在检索和浏览中思考并提出解决问题的新的思想和方法的过程。现有的DDMS难以快速为用户提供所需要的知识。从计算机学科的角度来看,造成这种现象的原因在于目前对信息、知识的表示和组织研究的不足。上述问题的研究进展,将会极大提高现有知识管理系统的工作效率。鉴于知识管理的重要作用,该研究具有重大的理论和应用价值。 本文重点对DDMS的索引系统和人机交互等方面作了深入研究,研究如何为检索者提供认知帮助以提高知识检索的质量和效率。主要工作有:针对目前DDMS缺乏语义支持的现状,提出了基于语义的文档索引系统的构建算法,并设计编写了原型系统,通过原型系统对所提出的理论和模型给出了实验验证;提出了针对DDMS中内容相近的学术文档的多种排序方法,并作了实验分析;设计了基于自然语言的智能人机交互界面,通过自然语言对话的方式渐进的了解用户的检索需求,帮助用户找到最符合检索目标的数字文档,并通过原型系统进行了验证;给出了迭代式的知识检索算法,用户通过迭代检索实时修正、更新检索目标,使DDMS给出的检索结果不断贴近用户的最终目标并具有个性化特点。更多还原

【Abstract】 In recent years, the power of knowledge management (KM) is widely recognized. Most enterprises consider that their continued survival in industry mainly depends on the successful implementation of KM. In terms of computer science views, KM is the methodology which is based on Internet and information technology, help enterprises manage knowledge resources systematically, locate expertise in organizations, build expert nets for teamwork and access special knowledge conveniently for better decision-making, etc. KM is meant to promote ongoing business success through a formal, structured initiative to improve the creation, collaboration in an organization.Nowadays, the research on knowledge management is in the exploring period. It has many problems in KM systems. The same situations are in knowledge retrieval of digital document management systems (DDMS), which is a part of KM system. In knowledge retrieval process, the main purpose of users is not to look for special data, but to learn or think through studying a set of relevant artifacts across several different application domains. Furthermore, the things that can stimulate or inspire them to generate new ideas in problem solving or let them think in number of parallel ways. At present, many established DDMSs often suffer from non-use. The main reason is the deficiency of research on knowledge representation and organization. Once above issues resolved, the efficiencies of DDMSs are greatly improved on.In this paper, the index system and human-computer interaction (HCI) of DDMS, which provide cognitive supports for users to improve the quality andefficiency of knowledge retrieval are discussed. We propose semantic-based algorithms in order to construct the index system of DDMS. The methods are verified by the design of a prototype of DDMS. We also study the paper ranking based on the similarity of the profiles or the importance of papers. The effect of the rankings is evaluated by informal user study and the comparison experiments with the traditional ones. We integrate the techniques used in natural language processing and taxonomies to understand users’ searching requirement and locate the artifacts that users may need in DDMS. Moreover, an iterative and interactive knowledge retrieval algorithm is designed, which can learn about customers’ true characteristic intentions through iterative interactions.The proposed theory and models in the paper are verified by algorithm analysis and prototype experiments. It is proved that all kinds of methods proposed in the paper can provide better cognitive supports for users to improve the efficiency during the process of knowledge retrieval.更多还原