【作者】 吴一占；

【导师】 马静；

【作者基本信息】 南京航空航天大学 ， 管理科学与工程， 2011， 硕士

【摘要】 随着语义网概念的提出,作为实现语义网关键的本体技术受到了广泛的关注和研究,如今在知识表示、知识管理、知识共享、知识复用等多方面都有着广泛应用,成为实现信息知识化处理的最有效方法之一。然而,人工维护本体难度过大、成本过高,极大地限制着本体技术向更广阔领域的运用与推广。现有的本体进化技术虽提出了不少解决方案,但是因为弱于挖掘本体中概念间的关联,依旧没有很好地解决问题。本文首先依据Web信息获取分析系统的应用需求,结合本体的构成要素,分析了该领域本体进化的目标;并围绕着实现本体进化,对现有Web挖掘技术进行了分析与研究。之后,本文设计了基于Web知识关联挖掘的本体进化方案,将实现的技术方案分解成流程框架与实现算法两部分分别进行讨论。在流程框架设计上,本文提出了本体进化系统与Web信息抽取挖掘系统相互结合,相互推进的流程框架,实现在循环进行的Web信息抽取挖掘的过程中完成本体的进化;在实现算法上,本文结合Web信息特征,以本体学习更准确、更高效为目标,改进和设计了各步骤的具体实现算法。最后,本文依据设计的理论框架开发了原型系统,验证了本文设计的本体进化方案稳定有效,一定程度解决了人工维护本体困难的问题,达到了本文研究的目标。本文有如下创新点:1、本文设计了Web知识关联挖掘算法策略,通过该策略处理可以获得实体间的关联关系,从而解决现有本体进化对于处理实体关联不足的问题。2、将Web信息抽取挖掘与本体进化相结合,设计了闭环相互优化的系统框架,同时满足了信息抽取挖掘系统和本体进化两方面的需求。3、针对Web信息抽取挖掘的应用特性,在处理流程上提出了改进算法,采用基于贝叶斯过滤的文本相关性判断算法、引入动态内容识别技术等,使本体进化更高效、更准确。更多还原

【Abstract】 With the proposed concept of the Semantic Web, as a key of the Semantic Web, ontology technology has been widely concerned and studied with a wide application in the knowledge representation, knowledge management, knowledge sharing and knowledge reuse fields. Ontology has been one of the preferred techniques in information knowledge solution. However, the artificial maintenance of ontology is too difficult and too expensive, which greatly limits a broader use and promotion of ontology technology. Although the current ontology evolution made a lot of solutions, it is weaker than mining the association between concepts of ontology and is still not well to solve the problem.Firstly, based on the application requirements of Web information retrieval and analysis system, combined with elements of ontology, the thesis analyzes the target of the evolution of domain ontology and does research on the existing Web mining techniques around the ontology evolution. Secondly, it designs Web knowledge association mining-based ontology evolution programs, divides finished technical solution into two parts, process framework and algorithm for discussion. In the design of process framework, the thesis proposes ontology evolution system and Web information extraction and mining system combined with each other to promote in the process framework and completes ontology evolution in the loop process of the Web information extraction and mining. Binding characteristics of Web information and with the target of more accurate and efficient ontology learning, the thesis improves and designs the specific algorithm in various steps. Finally, according to the designed theoretical framework, it develops a prototype system to verify that the evolution program is steady and effective, solving the difficult problem of artificial maintenance of ontology to some extent and reaching the goal of this study.This thesis has the following innovations: Firstly, this study designs algorithm strategy for Web knowledge association mining to obtain the relationship between the entities and solve the problems that the existing ontology evolution is inadequate for dealing with entity association. Secondly, combining the Web information extraction and mining and ontology evolution, it designs the closed-loop optimization system framework to meet the demand for both the information extraction and mining system and ontology evolution. Finally, aiming at the application characteristics of Web information extraction and mining, it proposes improved algorithm in processing flow and evolved ontology evolution more efficient and more accurate with the algorithm about the relevance of the text judgement based on Bayesian filtering and with the introduction of dynamic content recognition.更多还原