【作者】 葛继科；

【导师】 邱玉辉；

【作者基本信息】 西南大学 ， 基础心理学， 2009， 博士

【摘要】 服务作为一种自治的、开放的以及与平台无关的网络化构件,可以使分布式应用具有更好的复用性、灵活性和可增长性。作为现代服务科学的奠基石,服务计算已成为一项桥接商业服务与信息技术服务的跨学科的科学技术。一个IT支持的商业服务具有两个典型的特征:服务操作模型和服务费用模型。服务操作模型定义了服务如何被发现和调度的问题,包括服务建模、服务创建、服务发现、服务组合、服务提供,服务管理等,是服务计算研究中的重要课题;服务费用模型声明了被调度服务的费用如何的问题。语义网格将语义Web为代表的语义技术和以网格计算为代表的体系架构技术结合起来,对信息和服务进行了较好的定义,可以更好地实现计算机与人的协作。语义网格作为一个分布式、异构的开放式系统,具有高度的自治性,用来提供灵活的协作和大规模的计算,为资源的有效共享和高效处理提供直接支持。语义网格环境中的服务发现是一个重要的研究课题,它用来解决一个服务需求者如何发现解决特定问题的资源或服务,以及一个服务提供者如何使得服务需求者注意到他提供的服务。语义网格环境中的服务发现是在不同虚拟社区提供的异构的、分布的和共享的资源上进行的,服务发现的核心问题是服务匹配。服务匹配的最终目标是发现不同服务之间的相似性。两个服务越相似,则它们之间越匹配;反之,两个服务越不相似,则它们之间越不匹配。服务匹配在服务发现及调度过程中具有举足轻重的作用,如果缺乏有效的服务匹配策略,就不能为用户提供满卷的服务。统一描述、发现和集成(UDDI)作为一个公共的服务注册中心,为服务的发布与发现提供了一种高效、灵活和可扩展的机制。但是,它只是基于关键字的匹配,这种匹配方式既不能区分语法不同但语义相同的信息,也不能区分语法相同但语义不同的信息,因此,很难提供基于语义的服务匹配。同时,基于关键字的匹配也导致了服务匹配的查准率较低,其中一个可能的原因是检索的关键字从服务描述上看只是语义相似而语法不同;另一个可能的原因是检索的关键字从服务描述上看只是语法等价而语义不同。并且,这类方法也不能完全获取用户检索的语义,因为它们没有考虑关键字之间的语义关联关系。对该问题的一种可行的解决方案是使用基于语义的服务匹配,如果能够从语义上解决服务匹配的问题,将会极大地提高服务匹配的成功率及信息共享的程度。因此,对服务语义匹配的研究具有重要的理论意义和一定的实用价值。在服务匹配过程中,借助具有语义表述能力的知识表示模型,如本体及资源空间模型等,可以在一定程度上实现服务的语义匹配。语义匹配机制可以执行更加灵活的服务发现过程,允许在数值上进行推理而不仅仪是基于类型的推理。此外,他用语义还允许执行包含推理,这意味着服服务匹配过程不再是仪局限于服务名称的匹配。因此,执行服务的语义匹配可以发现基于关键字匹配方法不能发现的服务。通过对国内外研究现状的分析,在已有研究工作的基础上,本论文结合语义网格、语义Web及Web服务本体语言OWL-S,基于本体和资源空间模型理论,对语义网格环境中的服务匹配进行了相关研究。论文的主要研究工作和创新点体现在以下几个方面:(1)基于本体层次结构和向量空间模型,提出了一种用于服务语义匹配的扩展的余弦相似度度量方法发现服务实体之间的相似性从而更好地实现协同服务已成为服务匹配研究的一个热点问题。对服务实体相似性进行度量的传统方法主要是基于实体交集进行的,这种基于变集的度量方法不能准确地捕获特定领域的实体相似性。在当前研究的基础上,基于本体层次结构和向量空间模型,提出了一种用于服务语义匹配的扩展的余弦相似度度量方法。该方法能够在语义层次上获取更加符合人类直观认识的不同服务间的相似度,为实现基于语义的服务匹配提供了一种可供参考的方法。通过与传统相似度度量方法的对比,实验结果证实了该方法具有较高的精确度。以认知心理学的理论研究为依据,通过将该方法的计算结果与被调查人群的心理评价进行对比,验证了本文所提方法与人们直观认识的符合程度。(2)提出了一种基于语义距离的服务相似度度量方法为了实现服务请求与服务或者异构服务之间的语义匹配,关键问题是要找到服务实体之间的语义相似度。服务的语义相似度与服务实体间的语义距离以及服务实体所包含的子集等信息有关,此外,语义距离与语义相似度之间也存在着密切的联系。针对目前在解决服务匹配过程中对服务实体语义相似度度量研究的不足,以信息论的观点为基础,结合本体的特点,提出了一种基于语义距离的服务相似度度量方法。该方法综合考虑了概念之间的继承关系以及概念在本体层次结构中所处的位置对相似度的影响,有助于理准确地模拟客观世界的原貌,计算出的相似度更加合理。另外,针对概念集合组成的服务如何进行语义相似度度量的问题进行了深入研究,提出了概念集合匹配过程中的最大语义相似度度量方法、最小语义相似度度量方法、平均语义相似度度量方法和加权语义相似度度量方法,论证了概念集合加权语义相似度度量方法的合理性。最后,通过与不同相似度度量方法的对比,验证了本文方法的有效性。(3)提出了一种基于IOPE描述的服务功能匹配方法在OWL-S服务匹配中,不仅要考虑服务非功能描述信息的匹配,如服务名称、服务描述等,还应该研究服务功能描述信息的匹配,如服务的输入、输出、前提条件和预期的效果等。针对目前在解决服务匹配过程中对服务功能信息匹配研究的不足,结合描述逻辑,提出了一种基于IOPE描述的服务功能匹配方法。首先,对服务输入、输出、前提条件和效果等功能参数分别进行语义相似度匹配,然后,再计算出服务功能匹配的加权全局相似度,从而实现了服务功能的语义匹配。通过对权重的按需设置,使得该方法具有较高的灵活性。该方法是在服务匹配过程中对服务功能信息匹配研究的一个有益的尝试。通过对本文的方法进行定性分析和定量评价,论证了该方法的可行性。(4)提出了一种基于资源空间模型的服务相似度度量方法通过对资源内容进行分类,资源空间模型是一个规范、存储、管理和定位网络资源的语义数据模型,它采用多维资源空间的方式组织资源,支持有效的资源管理。随着资源空间模型研究和应用的不断深入,基于资源空间模型的服务匹配成为一个值得研究的课题。基于资源空间模型的服务匹配的实质是不同资源空间模型之间的相似度度量。根据资源空间模型的特点,提出了一种基于资源空间模型的服务相似度度量方法。通过对坐标相似度、轴相似度和资源空间相似度的计算,从而实现了基于资源空间模型的相似度度量。实验结果证明本文给出的方法在解决资源空间模型的匹配问题上是可行的。论文最后对研究工作进行了总结,提出了今后进一步的研究方向。本论文在基于语义的服务匹配方面所做的工作虽然具有一定的理论意义和潜在的实用价值,但是这些研究工作只足整个服务计算研究中的一小部分,作者将在现有的研究基础上进一步做深入的研究。更多还原

【Abstract】 Service as an autonomous, open and platform-independent network-based component, it makes the distributed application systems with better reusability, flexibility and growth. As the foundations of modern service science, service computing has become a cross-discipline subject that covers the science and technology of bridging the gap between Business Services and IT (Information Technology) Services. An IT-enabled business service is typically characterized by two features: its service operation model and its service charge model. A service operation model defines how the service is to be discovered and delivered, including services modeling, services creation, services discovery, services composition, service delivery, services management, et, al; a service charge model specifies how the delivered service is to be charged.Semantic Grid combines semantic Web with Grid computing technologies, and it is an extension of the current Grid technique, in which information and services are given well-defined meaning, better enabling computer and people to work in cooperation. Semantic Grid is a distributed, heterogeneous and open system, with a high degree of automation, which supports flexible collaboration and computation on a global scale. Semantic Grid makes resources be effectively sharing and efficiently processing.Service discovery in Semantic Grid environment is a fundamental research issues in answering the questions of how a service requester finds the services needed to solve its particular problem and how a service provider makes potential service requesters aware of the services it can offer. Service discovery defines a process for locating service providers and retrieving service descriptions. The problem of service discovery arises through the heterogeneity, distribution and sharing of the resources/services proposed by different virtual communities in Semantic Grid. At the heart of the service discovery is the concept of service matching. The ultimate goal of service matching is finding the similarity between different services, The more similar the two services, the more matching between them. Vice versa, the less similar the two services, the less matching between them.Service matching is a key role in the processing of service discovery and delivery. It can not provide satisfied service to users if we can not implement service matching strategy effectively. UDDI (Universal Description, Discovery, and Integration) is a public registry of published services, and it provides an efficient, flexible and extensible mechanism for service publication and discovery. However, it is a keywords-based matching strategy, and it can not provide semantic-based service matching. This, in majority of the cases, leads to low precision of the retrieved services. The reason might be that the query keywords are semantically similar but syntactically different from the terms in service descriptions. Another reason is that the query keywords might be syntactically equivalent but semantically different from the terms in the service description. Another problem with keyword-based service matching is that they cannot completely capture the semantics of users’ queries, because they do not consider the relationship between the keywords semantically. One feasible solution for this problem is to use semantic-based service matching method. If we can solve service matching problems semantically, it will greatly improve the success rate of service matching and advance the extent of information sharing. Therefore, the research on semantic matching of services has important theoretical significance and a certain degree of practical value.For realizing semantic-based service matching, a more ideal approach is to use the knowledge representation model with semantic expressing capability, such as ontology, resource space model, in the processing of service matching. Semantic matching mechanism allows a powerful and flexible service discovery process as it uses semantic service descriptions. Using semantics allows to reason on values which is not only based on type reasoning, it furthermore allows subsumption reasoning. This means that the service matching is very powerful as not only a service name matching is performed. Services which would have never been found with the "keywords" service matching methods can get discovered.On the basis of the state of the art of service matching at home and abroad, integrating with the features of Semantic Grid, Semantic Web, and Web Ontology Language for Service (OWL-S), we do some research on service matching problem in Semantic Grid environment based on ontology and resource space model.The major research works and contributions of the dissertation arc as follows:(1) An extensible cosine similarity measure for service matching semantically based on ontology hierarchical structure and vector space model.Finding similarity between service entities for realizing better cooperative service is an important issue in service matching domin. Entities being compared often are modeled as sets, with their similarity traditionally determined based on set intersection. Intersection-based measures do not accurately capture similarity between entities in certain domains. On the basis of the current research, integrating with ontology hierarchical structure and vector space model, we propose an extensible cosine similarity measure for service matching. This method can capture semantic similarity among services, and the captured semantic similarity is more in line with people’s intuition. The method provides a valuable reference for realizing the semantic-based service matching. In addition, we provide experimental comparison of our measure against traditional similarity measures. The results verify the better accuracy of our method. According to the theoretical research results of cognitive psychology, we also report on a user study that evaluate how well our method matches human intuition through comparing the results of our method with the psychological evaluation of investigated crowds.(2) A service similarity measure based on semantic distance.In order to accomplish semantic matching between service requests and services, or semantic matching among heterogeneous services, the important problem is that discovering the semantic similarity between service entities. The semantic similarity of services not only relate to the semantic distance between service entities, but also subsets of service entities. Another, the semantic distance and the semantic similarity have a close relationship. In order to resolve the shortage of the semantic similarity of service entities in service matching, we propose a service similarity measure method based on semantic distance on the basis of the view of information theory and the feature of ontology. The method synthetically considers the affects of inheritance relationship among concepts, as well as the position of concepts in ontology hierarchy. The acquired similarity is more reasonable, and it can more accurately simulate the original appearance of the real world. In addition, we study the semantic similarity of concept collections, and propose the maximum semantic similarity, minimum semantic similarity, average semantic similarity and weighted semantic similarity of concept collections, and demonstrate the rationality of the weighted semantic similarity measure of concept collections. At last, we provide some experimental comparison of our measure against other similarity measures. The results show how well our measure usefulness and feasibility.(3) An IOPE-based service functional matching method.In the processing of OWL-S based service matching, it is not only consider the matching of non-functional information of service, such as service name, service description, but also consider the matching of functional information of service, e.g. service inputs, service outputs, service preconditions and service effects. On the lack of the current research on the functional information matching in the processing of service matching, and combining with the Description Logic (DL), we propose an IOPE-based service functional matching method. In the method, firstly, we execute semantic matching of inputs, outputs, preconditions and effects separately. And then, we compute weighted overall similarity of service functional matching. This method can realize semantic matching based on service functional information. Through setting the weight on-demand, it makes the service matching process has high degree of flexibility. This method is a useful attempt of service functional matching in the service matching domain. At last, the evaluation of the method is done using a qualitative and a quantitative analysis, and the feasibility of this method is discussed.(4) A service similarity measure based on resource space model.A Resource Space Model (RSM) is a semantic data model for specifying, storing, managing and locating Web resources by appropriately classifying the contents of resources. Through multi-dimensional resource spaces, users can efficiently and effectively organize and manage Web resources. With the research and development of RSM, service matching based on RSM becomes a fundamental and valuable issue. The essence of RSM-based service matching is that the similarity measure between different RSMs. On the basis of the RSM’s features, we propose a RSM-based service similarity measure. Through computing the similarity of coordinates, axes and resource space separately, we fulfill the similarity measure of RSM. We evaluate our method on experimental data sets, and report empirically the strength of our approach.Finally, the research works in the dissertation are summarized and the future works are presented. Although the research works of semantic-based service matching in this dissertation have some theoretical significance and potential practical value, the research works are only a small part of the whole service computing research. We will do further study on the basis of the current researches.更多还原