【作者】 陈涛；

【导师】 肖侬；

【作者基本信息】 国防科学技术大学 ， 计算机科学与技术， 2006， 硕士

【摘要】 随着网格和语义网技术的飞速发展,语义网格应运而生。语义网格是当今网格的一种扩展,对服务进行语义的形式化描述来更好的促进计算机和人的合作。语义网格将网格上的资源进行更好的形式化描述,使计算机尽可能取代人进行信息处理,让传统网格更智能化。它是一个开放的系统,有着高度的自治性来提供灵活的协作和大规模的计算。语义网格将为资源的有效共享和高效处理提供直接支持。资源以服务的形式呈现出来,语义网格中服务的发现是个非常有挑战性的问题。本文设计了语义网格在资源共享和服务中应用的体系结构以及语义服务发布和发现模型;对经典的弹性匹配算法进行了改进,提出了服务的精确匹配算法;针对气象领域的需求,为气象建立本体库,在此基础上设计和实现了基于语义的服务发布和发现系统。语义服务发布和发现系统根据输入输出参数集合中本体概念的语义相似度来获得服务的语义相似度,服务发现的查全率和查准率有很大的提高;同时支持网格工作流的创建,为抽象工作流中的每个抽象服务找到具体的服务。本文研究的内容主要包括以下几个方面:1.提出了语义网格在资源共享和服务中应用的体系结构。本文采用基于网格工作流的体系结构,根据用户需求,灵活地分配网格资源,对其进行动态绑定。针对气象领域的特点,本文给出了语义网格在气象领域的实例研究。2.提出了语义服务的发布和发现模型。Web services的发布和发现模型没有包含服务的语义信息,服务发现时只是基于关键字的匹配。基于关键字的匹配使得查全率和查准率都比较低。本文对UDDI注册中心进行改进,使用本体对服务进行语义标注,进行语义服务的发布;服务发现时,使用服务的精确匹配算法进行语义推理匹配。3.提出了服务的精确匹配算法。它对经典的弹性匹配算法进行改进,弥补其不足。经典的弹性匹配算法将服务匹配分为四类,但是没有对每个类别内部的匹配程度进行刻画。本文借鉴基于语义距离和基于特征的相似度模型,根据输入输出参数集合中本体概念的语义相似度来获得服务的语义相似度。4.在语义服务发布和发现模型的基础上,我们针对气象领域的特点,为气象建立本体库,设计和实现语义服务发布和发现系统。最后分析说明本文提出的语义服务发布和发现模型与UDDI相比,其查全率和查准率有很大提高。更多还原

【Abstract】 With the rapid development of grid and semantic web technologies, semantic grid is emerging. Semantic Grid can be described as an extension of the current Grid in which information and services are given well-defined meaning, better enabling the cooperation of computers and people. The resources in semantic grid environment are described formally and computer process the information instead of human, making the traditional grid more intelligent. Semantic Grid is characterized by an open system, with a high degree of automation, which supports flexible collaboration and computation on a global scale. Semantic Grid makes resources be effectively shared and efficiently managed. Resources are exposed as services in Semantic Grid, and then the discovery of services is very challenging.In this paper, the architecture of sharing resources and services in semantic grid environment is presented, a model of semantic service publication and implementation is proposed and a precise matching algorithm which modifies the classical flexible matching algorithm is given, based on which the semantic services publication and discovery system is designed and implemented. This system supports semantic services publication and discovery, computes the semantic similarity of services by synthesizing the semantic similarity of inputs and outputs according to semantic similarity between concepts. The recall and precision of service discovery is greatly improved. Moreover, the system assists the creation of grid workflow and searches the concrete service for each abstract service.The main contents of our research are as follows:Firstly, the architecture of sharing resources and services in semantic grid environment is presented, which is grid workflow-oriented so that grid resources can be allocated flexibly and bound dynamically.Secondly, the model of semantic service publication and discovery is proposed. The Web Services model doesn’t refer to the semantic information of services and allows only a keyword-based search of services. Since search in UDDI is restricted to keyword matching and no form of inference or flexible match between keywords can be performed, recall and precision is very low in keyword-based search. In order to improve UDDI in Web Services, we annotate services making use of ontology by the time of publishing services, and then users can search them according to precise matching algorithm.Thirdly, the precise matching algorithm of service is proposed, which modifies the flexible matching algorithm to make up its faults. Service matching in flexible matching algorithm falls into four categories. It doesn’t compute the matching degree in each category. In order to resolve this problem, we compute the semantic similarity of service by synthesizing the semantic similarity of inputs and outputs according to semantic similarity between concepts.Lastly, we create meteorological ontology according to meteorological characteristics and implement the semantic service publication and discovery system based on the model of semantic service publication and discovery. Finally, we analyse the system and compare our model with UDDI.更多还原