【作者】 王莉；

【导师】 余雪丽；

【作者基本信息】 太原理工大学 ， 计算机应用技术， 2010， 博士

【摘要】 网络资源的极大丰富,为提升资源共享协同服务质量提供了可能;但是,网络资源数量和种类飞速增长的同时,资源管理和使用的复杂性也在增加。网络资源互联互通的通信机制是各种共享协同应用的基础核心架构,快速准确定位所需资源成为当前网络应用中一件非平凡的功能需求。新机遇总是和新需求结伴而生,技术的进步和丰富的网络资源数据及其可得性,为高效定位网络资源带来了可行途径。本论文以促进Web资源互联互通的共享为目标,首先对网络信息资源的不同表示形态进行了分析,从可计算角度,针对网络资源表示特点,根据知识的定义,给出一种知识通信架构及特征,其内涵是利用知识工程技术,基于社区发现和自组织构造,进行知识资源的传递和共享。接着围绕本论文所提出的知识通信基础设施中的三个核心问题:异构资源互通、资源虚拟组织管理、智能通信协议,以语义计算、社会计算为关键技术,以知晓内容和上下文为主线,进行深入研究和分析,提出一系列解决方法,归纳为以下几点,也是本文的主要创新点和贡献:1.着眼于资源语义模型和语义相关计算两方面进行异构资源语义共享研究。对异构资源共享中资源互理解机制进行深入分析,从资源表现形式、资源语义关联和资源协作关系等多个维度建立了网络资源的领域本体语义模型,并建立了包含认知过程类和知识协作过程类的知识协作过程本体,对用户的认知过程和资源协作过程规则进行描述,在资源描述层次支持协作式的智能服务推荐。针对W3C推荐的OWL本体描述语言,从描述逻辑的包含关系入手,对概念间存在的多种关系属性进行分析,建立包含关系的推理规则,基于此规则,缩小概念相似关系的计算空间,设计算法计算概念间相似关系;然后再返回到原图,设计相应规则,建立基于图结构的概念相关关系的计算方法。2.社区结构是社会网络中的一个重要结构特征,也是Web资源分布的一个重要特征。本文从宏观层次的网络资源社区拓扑发现和微观层次网络资源自组织社区构造两个不同角度进行网络资源组织的研究。当前重叠社区研究中普遍存在的问题有,缺少合适的社区质量度量体系,无法确定社区合适粒度,计算复杂性等问题。本论文以社区聚集系数和社区间的重叠度为因式建立了新的重叠社区结构质量评价函数,该评价方法可以同时评价重叠与非重叠社区结构,还能通过有效控制社区间的重叠程度达到优化重叠社区粒度的目的。与当前主流重叠社区发现算法从节点间直接连接关系分析入手不同,本论文以节点上下文间的关系为观察对象,提出一种基于节点间共享邻居关系的分层重叠社区发现算法NHOC。该算法不仅能发现重叠社区,也能发现非重叠社区,而且在社区融合过程中,算法会记录下社区分层结构。实验表明,与当前被看好的重叠社区发现算法CONGA相比,无论是在发现重叠社区,还是发现非重叠社区,NHOC得到的社区结构都比CONGA的结果更接近实际网络的社区结构。复杂社会网络中更多存在的是动态演化社区。论文不同于当前基于优化的动态社区发现算法,提出一种计算历史信息与瞬时信息的、多图转换的动态社区发现算法NDCD。该算法对瞬时观察网络为无权图(论文中表示为0-1图)和带权图分别进行了设计。其中,在计算0-1图中节点邻居影响时,把0-1图转化为带权图,考虑历史信息随时间失效的情况,计算瞬时综合信息,得到瞬时社区拓扑,然后对相邻时间社区拓扑的相似性进行计算,发现稳定社区。实验表明,无论观察网络是无权网络还是带权网络、观察数据是有噪音还是无噪音,NDCD都能较为准确的发现稳定社区结构。网络资源的动态开放性以及资源背后主体主动性在资源上的体现,使得资源之间呈现的是一种对等结构P2P。资源组织定位、质量管理是影响P2P架构广泛应用的瓶颈,当前这两方面的研究一般是分开进行。而实际上,通过建立虚拟语义社区把无结构P2P映射为有内容含义的结构化的网络架构,在社区组织方式的基础上,对搭便车、恶意行为、错误信息资源等进行识别和遏制,是一种比在P2P物理层上直接建立抑制机制更灵活的方案,可以同时实现资源组织和质量管理目的。基于此理念,本论文提出了一种P2P语义社区模型及其构造过程,设计了自主节点和自主语义社区模型,建立了基于节点本地视图的领域信任和被信任评价体系,使得节点可以通过收集通信历史记录,不断更新本地网络拓扑视图,建立多语义的网络拓扑知识图,并进行节点间协作,推选出质量最高的专家节点,然后以此专家节点为语义社区种子,建立专家服务并管理下的自组织语义社区。本文所建立的以专家为核心的自主语义社区模型,支持非社区节点、社区普通节点、社区专家不同角色的路由表动态更新,不仅能够有效增强快速定位资源的能力,而且在一定程度上抑制了P2P中低劣资源的传播以及搭便车行为泛滥。3.基于本体技术深入研究了感知内容和上下文的知识通信协议,以言语行为理论为支撑,建立包含了内容content、上下文context、动作act的协议本体,并基于事件演算,建立适应性协议编解码规则,在增强语义内容通信能力的同时优化通信负载,达到动态适应性的知识通信目的。最后构建了知识通信的原型系统——高等教育资源语义共享平台,对本文内容进行综合应用,系统运行结果表明,知识通信是促进web资源共享协同的一种可行且有效的基础设施。更多还原

【Abstract】 More and more resource exists on Internet and this makes possible to improve the quality of resource sharing and task collaboration on Internet. But, the increase of resource amount and categories adds the complexity of managing and utilizing resource. How to find the required resource, how to transfer resource quickly and accurately, these are nontrivial functions in current web applications. Semantic technology and social computing provide some new ways to realize web resource shared on application layer.Base on semantic computing and social computing, this dissertation aims at providing some new methods to enhance Web resource sharing and interoperability. We firstly analyze web resource’s characteristics and put forward the concept of web knowledge resource. This concept explains what web knowledge resource is and why most web resource can be considered as knowledge resource. Then to realize the widerly, high quality web resource sharing, a knowledge communication framework is projected, which is to deliver web knowledge resource by knowledge engineering and communication technology. In knowledge communication framework, the mechanism for connecting different resource by semantic content, effectively organizing and managing resource, intelligent communication protocol, constitute the infrastructure of knowledge communication. Focusing on the three core problems, we introduce some new methods and Content awareness and context awareness are these methods’medium objectives and resource widely shared is their final goal. The main contributions and innovations of our work include the following:①After deeply analyzing the mechanism for resource interoperability, we build multi-dimensions web resource ontology model RSM for resource semantic interaction. RSM includes metadata-based unification resource representation, semantic relationship and resource collaboration. In RSM, knowledge collaboration ontology is set up that is composed of cognize process class and knowledge collaboration class. They can respectively describe the user’s cognizing process and resource collaboration process. RSM could support high quality resource location and intelligent recommend on WWW.Another important notion about semantic interaction between heterogeneous resources is semantic relativity evaluation. We focus on OWL that is recommended by W3C and introduce an easy-to-use, flexible algorithm for computing semantic relativity between concepts in OWL, based on ontology structure and concepts’semantic relation, which does not need other lexical base such as WordNet, nor machine learning from quite large quantity of individuals. It just occurs in a domain ontology base. Because in DL, subsumption is an important relation, the detailed subsumption theories and computing lemma are given. According to these principles, we can deduce the subsumption subgraph and calculate their semantic similarity and then obtain their relativity degree by original ontology graph.②Community is one important structure in social network and web resource network. And it is one key structure for organizing web resource. We research community detection algorithms in macro view and self-organizing community formation in micro view. The work in macro view could help designing suitable micro self-organizing topology and the actions in micro world would influence the macro network structure.In macro view, we study overlapping community and dynamic community detection algorithms. There are some common problems in overlapping community research, such as lacking of suitable evaluation function, incapable of deciding suitable granularity community, computing performance and so on. This dissertation build a new overlapping community evaluation method that computes both community clustering degree and the overlapping degree. This function could calculate the structure quality for both disjoint community and overlapping community.Moreover. It can optimize the community granularity by controlling the overlapping degree.One new algorithm NHOC is proposed to detect overlapping or disjoint community. The difference between NHOC and current published algorithms is that NHOC utilizes the relationship of nodes’sharing neighbors, not nodes’direct interaction. Comparing it with overlapping algorithm CONGA, experiments show it can find more accurate community structure than CONGA, whether the real network has disjoint or overlapping communities.Community structure will change with time because social network is dynamic evolving. To find stable community in large social datasets that maybe have noise data, is one key problem. Different from current optimized dynamic community model algorithms, we introduce a graph transfer-based algorithm NDCD that computes history information and snapshot information. NDCD could detect community structure when the observed snapshot network is 0-1 (unweighted graph) graph or weighted graph. Experiments shows NDCD could discover the stable community whenever observed graph is 0-1 graph or weighted graph.In micro view, for autonomous community, we consider peer to peer modal as background because its sociality and widely applications on Internet. An expert-driven multi-semantic self-organized P2P semantic community model is built to resolve the two principle problems of resource organizing and management, which are how to possess enormous number of high quality shared resource, how to find the resource quickly and accurately. We set up Autonomic Semantic Community (ASC) model, design local domain-related trust function and self-evaluation function that takes requirement and owned resource two aspects into account to value one autonomous peer (AP). We also put forward a procedure to collaboratively elect expert APe and a Self-organized community formation algorithm --SoFA. In SoFA, APs gather its communication information and evaluate itself and other APs locally, build and dynamically adjust local topology, elect the domain expert APe and then an expert-driven ASC is established. Furthermore, we describe the dynamic changing of peers’routing table before or after that ASCs are formed. Experiment shows SoFA could not only adaptively optimize network to improve the efficiency of service discovery and routing, but also base on the synthesized trust degree to discriminate the low quality service peers and to keep them within limits.③Inspired by human’s communication way, we proposed a kind of content and context awareness protocol (abbr. KC2A2P) base on knowledge to satisfy the intelligent knowledge communication. A common shared protocol ontology is built that includes content, context and act in order to not only guide the different protocol terms mapping but also provide an intelligent protocol encoding framework. To avoid unnecessary communication load, event calculus is used to build the axiom of communication cognition and three basic encoding rules are made that can adjust the message entity base on the application context. Furthermore, some experiments are given to show KC2A2P would improve the meaning delivery and even help to build trust relationship.Finally a knowledge communication test system“higher education resource sharing platform”is introduced and some tests results show that knowledge communication is one effective infrastructure for web resource sharing.更多还原