【作者】 王星；

【导师】 马宗民；

【作者基本信息】 东北大学 ， 计算机应用技术， 2011， 博士

【摘要】 语义Web是当前Web的延伸,它赋予Web资源机器可理解的语义,从而使计算机能够更好地与人协同工作。语义Web的目标之一是推理Web上广泛存在的知识和数据,因此如何表示这些知识和数据以及如何利用它们进行推理成为语义Web的重要研究课题之一。作为知识表示和推理的重要工具,规则具有表达能力强、简单、易理解以及所需存储空间小等特征,这些特征使得规则在人工智能等许多领域发挥着重要作用,并已被应用于标记电子商务和语义Web的构建。随着规则在语义Web中的发展,各种规则语言相继出现。但是,这些规则语言无法表示语义Web中大量存在的模糊知识。近年来,Zadeh提出的模糊集被用于表示语义Web的模糊性,语义Web模糊规则语言应运而生。除了模糊性,知识的不完全性还带来了语义Web规则的非单调性特征,如何表示和推理具有模糊性和非单调性的知识被提上语义Web的议事日程。此外,随着各种具有异构性的模糊非单调规则语言和规则系统在Web服务等领域的广泛应用,模糊规则的互换成为一个亟待解决的重要问题。有关语义Web模糊规则语言和规则互换的研究已经取得了一些成果,与此同时,规则互换格式RIF (Rule Interchange Format)已经成为了W3C (WWW Consortium)规则互换方面的推荐标准。但是应当指出的是,已经出现的语义Web模糊规则语言不能同时表示和推理模糊非单调的知识,目前的规则互换不是以RIF为中心,有关模糊规则互换及其标准格式的研究几乎是空白。从语义Web模糊知识的表示出发,基于模糊集理论、可能性分布理论和Vague集理论,本文提出了能表示和推理模糊性和非单调性的语义Web模糊规则语言,并对RIF进行模糊化处理,提出了模糊规则互换格式f-RIF (fuzzy RIF),在此基础上构建了以f-RIF为中心的模糊规则互换结构。本文的创新性研究成果包括以下几个方面：(1)针对经典规则语言不能表示语义Web模糊及非单调知识的事实,提出了四种语义Web模糊规则语言。首先,把否定和否定即失败引入模糊规则语言,提出了能表示非单调模糊知识的规则语言f-NSWRL-(fuzzy NonmonotonicSemantic Web Rule Language);其次,在Vague集的基础上,提出了能够用区间来更精确表达模糊知识的规则语言Vague-SWRL,定义了概念二级权重,重点讨论了该语言的语义,并定义了该语言的XML Schema结构；再有,在充分吸收人工智能领域已有研究成果的基础上,把if-then规则引入语义Web,提出了规则语言f-SW-if-then-RL (fuzzy Semantic Web if-then Rule Language);最后,利用unless规则扩展了f-SW-if-then-RL,提出能同时表示模糊知识和非单调知识的规则语言f-SW-if-then-unless-RL (fuzzy Semantic Web if-then-unless Rule Language),并为其定义了语法、语义和XML Schema。(2)针对RIF不能表示和互换模糊规则的事实,提出了模糊规则互换格式f-RIF。首先,从可视化建模的角度定义了包括模糊项、模糊公式、模糊类、模糊属性、模糊约束和模糊规则蕴含等元素在内的f-RIF元模型；然后,基于给出的f-RIF的表示语法,构建了f-RIF模糊规则的UML Profile;接着,定义了f-RIF的XML Schema结构来表示f-RIF的具体语法；最后,基于模糊集和扩展了的真值集,定义了f-RIF的模型语义。(3)为解决异构系统之间模糊规则交互的问题,构建了模糊规则互换结构f-RIA (fuzzy Rule Interchange Architecture)。首先,以f-RIF元模型为中心并结合模糊规则语言f-SWRL、f-NSWRL、Vague-SWRL、RuleML、f-SW-if-then-RL和f-SW-if-then-unless-RL的元模型,构建了模糊规则互换结构f-RIA,实现了上述七种语言之间的抽象语法互换以及这七种语言中同种语言内部抽象语法和具体语法之间的互换；接下来,讨论了规则互换过程中出现的信息缺失问题并给出了相应的对策,以尽量减少信息缺失所带来的不利影响；最后,基于f-RIA,在Eclipse平台上设计实现了模糊规则互换原型系统f-RIA1.0。更多还原

【Abstract】 The Semantic Web is an extension of the current Web, in which Web resources are given computer-understandabale semantics, better enabling computers and people to work in cooperation. One of the declared goals of the Semantic Web is to reason about knowledge and data which are pervaded on the Web. Therefore, how to represent these knowledge and data and how to reason with them have become very important research topics of the Semantic Web. As an important tool for representation and reasoning of knowledge, rules possess many features including high expressiveness, terseness, easy understanding and requiring small memory space, which let rules play important roles in many areas including Artificial Intelligence, and have been applied to markup the constructions of E-Business and the Semantic Web.With the development of rules in the Semantic Web, rule languages came into being successively. However, these rule languages cannot represent much fuzzy knowledge existing in the Semantic Web. In recent years, the fuzzy set theory proposed by Zadeh is employed to represent the fuzziness of the Semantic Web, and thus fuzzy rule languages in the Semantic Web came into being. Besides fuzzniess, the Semantic Web rules also have another feature:nonmonotonicity, which is caused by the incompleteness of knowledge. How to represent and reason about these fuzzy and nonmonotonic knowledge has been firmly put on the agenda of the Semantic Web. In addition, with the widespread application of heterogeneous fuzzy nonmonotonic rule languages and rule systems in areas including Web Services, fuzzy rule interchange becomes an important issue to be solved.Researches on fuzzy rule languages and rule interchange in the Semantic Web have obtained some achievements, and at the same time, RIF (Rule interchange format) has become a recommendation of the W3C (WWW Consortium) on rule interchange. However, it should be pointed out that, the existing fuzzy rule languages in the Semantic Web are not capable of representing fuzzy and nonmonotonic knowledge simultaneously, the current rule interchange isn’t centerd on RIF, and there have been few efforts at investigating fuzzy rule interchange and its standard format.Proceeding from the representation of fuzzy knowledge in the Semantic Web, based on the fuzzy set theory, possibility distribution theory and vague set theory, this thesis proposes several Semantic Web fuzzy rule languages capable of representing and reasoning with fuzziness and nonmonotonicity, the RIF is fuzzified, fuzzy rule interchange format f-RIF (fuzzy RIF) is proposed, and based on the above, a fuzzy rule interchange architecture centered on f-RIF is constructed. The innovative contributions of this thesis are as follows.(ⅰ) Aiming at the fact that crisp rule languages don’t suffice to represent a great deal of fuzzy and nonmonotonic knowledge in the Semantic Web, four fuzzy rule languages are proposed. Firstly, negation and negation as failure are introduced into fuzzy rule languages, and f-NSWRL (fuzzy Nonmonotonic Semantic Web Rule Language) is proposed to represent nonmonotonic fuzzy knowledge; next, on the basis of the vague set theory, Vague-SWRL employing intervals to more precisely express fuzzy knowledge is proposed, a new concept second degree weight is presented, and the semantics and the XML Schema of Vague-SWRL are discussed and defined, respectively; then, based on research fruits in Artificial Intelligence, if-then rules are introduced into the Semantic Web, and fuzzy rule language f-SW-if-then-RL (fuzzy Semantic Web if-then Rule Language) is proposed; finally, combining unless rules with f-SW-if-then-RL, f-SW-if-then-unless-RL (fuzzy Semantic Web if-then-unless Rule Language) which can simutaneously represent fuzzy and nonmonotonic knowledge is proposed, and its syntax, semantics and XML Schema are defined,(ⅱ) Aiming at the fact that the RIF is not capable of representing and interchanging fuzzy rules, f-RIF, a fuzzy rule interchange format, is proposed. At first, the f-RIF metamodel including fuzzy terms, fuzzy formulas, fuzzy classes, fuzzy properties, fuzzy restrictions and fuzzy rule implications is defined from the view of visually modelling; then, based on the given f-RIF presentation syntax, a UML Profile for f-RIF fuzzy rules is also constructed; after that, the XML Schema of f-RIF is defined to represent the concrete syntax of f-RIF; at last, based on the fuzzy set theory and the extended set of truth values, the model-theoretic semantics of the f-RIF is defined.(ⅲ) To deal with the problem of communications of fuzzy rules among heterogeneous systems, a fuzzy interchange architecture f-RIA (fuzzy Rule Interchange Architecture) is constructed. Firstly, centered on the f-RIF metamodel and the metamodels of f-SWRL, f-NSWRL, Vague-SWRL, RuleML, f-SW-if-then-RL and f-SW-if-then-unless-RL, fuzzy rule interchange structure f-RIA is constructed, and abstract syntaxes transformations between the above seven languages and the transformations between abstract syntaxes and concrete syntaxes of the seven languages are implemented; next, the issue of information loss occurring in the process of rule interchange is discussed, and the corresponding remedial measures are given to reduce the adverse influences caused by information loss; finally, based on the f-RIA and on the Eclipse platform, a prototype system f-RIA1.0interchanging fuzzy rules is designed and implemented.更多还原