【作者】 王瑜；

【导师】 胡运发；

【作者基本信息】 复旦大学 ， 计算机软件与理论， 2004， 博士

【摘要】 随着信息技术的迅猛发展，人们所掌握的知识呈现出爆炸性增长的趋势。统计表明，每五年人们的知识容量就会翻一番，超过百分之九十的经济活动都与知识有关。为了适应知识社会的发展，更合理、有效地管理和利用知识，我们需要对知识本身及其活动规律展开研究，这也是知识工程领域的核心研究内容。作为知识工程领域的重要组成部分，知识度量、知识错误检测及知识融合等子课题不仅具有重要的理论研究价值，而且对实际应用项目也具有积极的指导意义。 本文结合科研项目背景中的实际需求，针对知识工程领域中的上述相关问题展开了研究。具体包括确定性及非确定性知识含量度量、知识推理过程中的错误检测及多源知识融合等方面。本文所完成的工作及取得的创新成果可主要归纳为以下四个方面： 1．确定性知识含量测度准则研究 通过对划分等价类个数、各等价类基数、各等价类包含对象这三方面因素的考察，本文归纳得出了确定性知识库的三种相似性约束关系：等价、等构及等势，并在此基础上给出了知识含量测度必须遵循的四条准则。基于此四条准则，本文证明了熵测度的合理性，并提出了一种新的基于分类的知识含量测度。除此之外，本文还围绕确定性知识含量度量的研究工作，就相关其他问题展开了讨论，其中包括：给出了知识库知识差异性的度量公式及范畴简化的度量公式，并从理论上证明确定性知识库的叠加操作不可以产生新的知识。 2．模糊性知识含量度量研究 将度量问题的研究对象由确定性知识进一步扩大到模糊性知识的范围。针对模糊集上常规运算符在运算后会损失知识含量的问题，本文依据分类的思想重新选择了分类交、分类并、分类差等运算符，并由此构造出了模糊集和模糊集簇的知识含量度量公式，并对其正确性进行了证明。 3．基于最小执行集的知识推理中错误检测研究 针对产生式推理模型中的一种可能导致推理不可结束的错误，本文在给定约束条件下对其产生机制进行了研究，通过引入结点及结点集的最小执行集等概念，将判定结点及推理模型是否可推理问题转换为计算结点的最小执行集的问题，提出了此类模型错误的检测算法。理论分析和试验结果均显示该算法可以完备地找到推理模型中所有此类错误。 4．基于映射函数的多源知识融合引擎研究与开发 针对目前基于XSLT技术的数据交换平台所存在的不足，本文提出并实现了摘要一个多源知识融合引擎。该引擎以映射函数作为基本的计算单元，通过映射函数的组合实现复杂的映射功能。该引擎弥补了现有系统源数据单一，编写复杂等缺点，可直接从多XML文档、关系数据库、Web页面提取数据，并提供可视化快速建模工具。该引擎目前已在上海市电子政务协同工作平台中投入实际运行。更多还原

【Abstract】 Along with the rapid development of information technology, knowledge in people’s pockets rushingly swell. According to the recent statistics, the column of people’s knowledge with which beyond 90% of economy actions now associate would double very five years. Not only for adapting to the impending development of knowledge society, but for managing and taking advantage of the power of knowledge, we must delve into the knowledge and it’s development laws deeply, which are the key research topics in the field of Knowledge Engineering nowadays. As the indispensable parts in this area, the subjects of the measurement, the error-detecting and the amalgamation of knowledge not only are unvaluable amongst the theoretical researches, but at the same time all make contribute to the applications of the practical projects greatly.The thesis addresses several key technical problems of knowledge engineering and its search based on the practical requirement from research project, which covers precise knowledge and fuzzy knowledge measurement, error-detecting in knowledge reasoning model and knowledge amalgamation. Major contributions of this thesis include:1.Research on the principle of measurement of precise knowledgeWith the consideration of three factors: the number of subsets, the cardinality of each subset and the elements in each subset, three relative constrains are conculded in this thesis, which are equivalence, equal-structure and equal-potential. Based on three constrains, this thesis brings up four principles by which the measure of knowledge capcity must abide. This thesis proves the rationality of Hent and brought an new measurement of knowledge capcity which is based on the thought of distinguishability. More over, the thesis provides the measurement of the difference between two knowledge bases, and proves theoretically that new knowledge can not be incubated through the union of precise knowledge bases.2.Research on the measurement of fuzzy knowledgeThe target of the research in this thesis includes not only precise knowledge but also fuzzy knowledge. Because normal fuzzy set operations may lose the knowledge in fuzzy sets, three extended operations are defined in this thesis based on the thought of distinguishabillty. The measurement of fuzzy knowledge in fuzzy set and fuzzy sets are constructed, and the correctness of the measurement is provided by the thesis.3.Research on error-detecting in knowledge reasoning model based on least execution set.Under some certain restrictions, the mechanism of one kind of error in reasoning model is discussed in this thesis, which could lead the model to hang up. By introducing some concepts about the execution and the least execution set of node and node set, Task about judging whether nodes and model are reasonable are converted into the task of calculating the least execution set of node, and the error detecting algorithm is provided in this thesis. Theoretical analysis and experiments show that all of this kind of error in model can be detected by this algorithm.4.Research and development of multi knowledge sources amalgamating engineComparing with deficiency of existing data exchanging plaform based on XSLT, a multi-knowledge sources based amalgamating engine is designed and realized in this thesis. The mapping function is treated as the basic computing unit in this engine, and the complex mapping operation can be accomplished by the composing of several mapping functions. This engine can extract data directly from multi-XML documents, relational database and Web pages, and provide a visual tool to build model quickly. Now the engine has been taken into effect as one of key components of Shanghai E-GOV Interoperable Platform.更多还原