Research On Service Semantic Link Network Based On Probalisitic Graphical Model

It is of importance for the service-oriented community to efficiently locate a reasonably small group of service candidates from the enormously services space and to automatically create a list of related services that match a given query. This fact calls for the development of methods for service automatic discovery to automatically creat a list with related services organised and their functionalities semantically described. And it is the key for automatic service search and collaboration in order for maximizing the utility of Web services by making them widely available to the community. The semantic model of Service Semantic Link Network (S-SLN) is employed to define the semantic association structure among Web services. S-SLN is the underlying semantic model for effectively implementing automatic Web service search and collaboration by a relationship dependency network which connects services with different types of relationships. The dissertation mainly focuses on the S-SLN discovery and S-SLN based service semantic relationship reasoning for service automatic recommendation. The main contributions of this dissertation are as follows:Because of the growing number of Web services becoming available, Web services are being made available to increase collaboration in distributed environment. However, discovering such resources to facilitate automatic services collaboration is a major bottleneck in this context. As a way to effectively collaborate between services. Web services need to be organized, their functionalities semantically described for discovering semantic associated service network among domain-specific Web services. This dissertation applies probabilistic graphical model to service network discovery. Markov network is evoked to describe the service relationship dependencies among services based on the joint probability distribution and a method is proposed for discovering S-SLN by virtue of transforming Markov network into directed graph structure.However, there are service semantic relationships with uncertainties in S-SLN. In addition to the explicit service semantic relationship, there are some other potential relationships in S-SLN. Semantic relationship reasoning with uncertainty in S-SLN is one of the challenges for S-SLN based advanced Web service application. To solve this problem, the dissertation presents an approach to service semantic relationship reasoning in terms of Markov Logic Network which is represented by combining first-order logic with probabilistic graphical models in a single representation. In this dissertation, the Service Markov Logic Network is constructed using available probability information and knowledge. And it is taken as a logical framework for semantic relationship reasoning with uncertainty to discover and predict implict service semantic relationship.As an advanced application of S-SLN, the dissertation proposes an approach to S-SLN based service automatic recommendation using spreading activation technique. Automatic service recommendation and navigation is a major objective of S-SLN, and we employ the spreading activation method as a scalable and effective solution for automatic service search and navigation and map related service nodes recommendation problem as a basic graph mining problem. These closely related service nodes linked by different relations form a subgraph of S-SLN, in which service nodes and links with different types and weight represent different service scenarios and relations to the maximum extent.The experimental results and analysis show that our proposed approaches are efficient, feasible and applicable.