A Similarity-Based Matching Framework For Semantic Web Services

With Web Services being widely deployed, a dramatic increase can be predicted in both the number and category of Web Services. Faced with such an immense group of services, the ability to perform accurate and effective discovery of services against user requirement become of great significance. In spite of the matured WSDL and UDDI based Web Services, their matchmaking mechanism is still kind of syntactic level, which is not sufficient to solve the above problem. That's the reason Semantic Web technologies were introduced into the Web Service community. Semantic Web Service was proposed to integrate Semantic Web and Web Service technologies to enable automation in service discovery, matching, invocation and monitoring.This thesis explains the role of Semantic Web technologies in the matching of Web Services by two typical matching scenarios of Semantic Web Services, and gives some analysis and discussion about matching issues within the Semantic Web Service and then proposes some major viewpoints: (1) Common ontology is the key to widely accurate matching; (2) Service matching should be divided into functionality matching and signature matching; (3) Functionality matching should be performed by the use of service category information, while input/output information is probably more appropriate to be taken as an exhibition of invocation method of a service. Base on the above viewpoints, the thesis proposes a similarity-based service matching framework.This work has a two-fold contribution.? In terms of service description, the framework takes up the Property Classifying technique for the representation of inputs and outputs of a service, which will improve the maximize the utility of the Profile Hierarchy approach.? In terms of service matching, the service category information and service inputs/outputs information are designed to match separately. The matching of service category is served as functionality matching, while the matching of inputs/outputs is served as signature matching. The matching results finally lead to an overall result based on their individual weights. The matching framework is designed to have great flexibility in that the weights can be tuned to adapt to different scenarios.