Implements The Self-Organizing, Retrieving And Recommending Of The Resources In P2P Networks To Support The Semantic Understanding

The rapid development of network technology in recent years makes network application much more important through our lives. However, the abundant resources in the network inevitably brought the new challenges in the network application while the scale of network is persistently expanding. The organization, searching, managing, recommendation and using of the diverse resources including CPU, storage, bandwidth, cache, files and services became the working focus of the information retrieval. The intrinsic flaws of the traditional network computing model (Client/Server model) makes the emerge of P2P model inevitable. P2P model is the typical representation of the new models which can overcome the disadvantages mentioned above through its self-organizing character, symmetrical character and the adhoc character. The three major characters make the organzing, retrieving, managing and using of the resources in P2P networks much more complicated which can not be implemented merely relying on the traditional techniques. How to effectively organize, retrieve, manage and use of the resources in P2P networks is the main content we will discuss in this paper.Overlay network is the most popular technique used to implement the goals mentioned above. Should they be "structured" or "unstructured"? Are they complementary to each other or competed to each other? This is the main problem we will solve in this paper. As a result we propse a new method to construct the overlay network which neither need to rely on the supporting topology of the real network, nor need to consume the huge network bandwidth to guarantee its search efficiency. A middleware TRM is propsed to construct the totally distributed P2P overlay network and to efficiently support the complexed semantic queries of the resources through its decentralized, dynamic, expanding and robust fault tolerant characters. TRM provides the ability of automatically constructing the resources based on ontology concepts to implement the self-organzing and dynamically clustering of the overlay networks (DOverlay) which further guaranteed the search radius of the semantic query. Resources in DOverlay is neither labeled by DHT nor labeld by key vector, but labeled by the character vector in ontology concepts. The search policy against DOverlay is named TRM_Search which can support the complex semantic query, while the resource fusing policy is called TRM_Evaluate which can synthetically evaluates the resources and make the decision.On the other hand, the recommending policies are not flexible enough to satisfy the requirements of the P2P network, such as adhoc character because of omitting the users' characteristics and the inaccurate reseaoning of the user model. This is the main reason to stimulate us to apply more professional user cognition characters in modeling users' requirements and reasoning users' actions under the context aware computing environment. After researching the context aware computing environment, users' cognition character which effecting their cognizing process is fund to be valuable in constructing the user profile model and providing users with the personalized services. In this paper, a new user model which sufficient considered the user's cognition character (Field Dependence/Independence) in context aware computing. This FD/I model can improve the accuracy in resource recommending in the adhoc P2P network. Of cause, only the user model can not finally provide the high quality personalized services without an effective user model reasoning mechanism. In order to improve the performance of user model reasoning, we provide a case computing policy which abstracts the key words from the scenes to construct an analog ontology concept trees to compute the similarity of the two scenes. The FD/I user mudel and the anolog ontology context based case reasoning policy can provide more accurate and more efficient resource recommending in the adhoc P2P environment which is proved through the simulating experiments.The research work and creations of this paper are including as follows:(1) Propose a middleware named TRM which can be used in the totally decentralized P2P applications through constructing the' self-organized and dynamic clustered resource overlay (in this paper is called DOverlay) and supporting the complexed semantic queries.(2) The DOverlay mentioned above is neither labeled by DHT, nor labeled by key word vector, but is labeled by character vector of the ontology concepts.(3) Propse a resource search policy TRM_Search to satisfy the complex semantic queries against DOverlay, and a result fusion policy TRM_Evaluate is proposed to synthetically evaluate the resources searched by TRM_Search.(4) A user model which sufficient considered users' cognition character FD/I is propoed to improve the efficiency of resource recommending in adhoc P2P environment.(5) A user model reasoning policy which is based on anolog ontology context in context aware computing is being proposed to finally improve the accuracy and efficiency of the resource recommending.