Semantic Discovery of Web Services through Social Learning

摘要

The increasing numbers of web services impose automatic discovery process in Service Oriented Architecture (SOA). But the existing SOA enables only syntactic discovery which produces coarse irrelevant results or sometimes no results. Different researches challenge this problem by introducing semantic discovery process in SOA to enable relevant and desired search results. These research outcomes cannot discover services efficiently which are created independently with different knowledge bases. To overcome these problems, a new architecture of SOA is proposed which incorporates a new adaptive technique called social learning that improves service provider's domain ontology from service consumer's concept contributions and thus eventually makes the service more semantically discoverable. The proposed architecture contains new similarity measure and automatic merging algorithms on weighted ontology. From mathematical reasoning it is induced that the proposed architecture reduces overlapping concepts and thus more relevant discovery results are ensured. To test the proposed architecture's performance, a prototype of Universal Description Discovery and Integration (UDDI) is implemented and a simulation is conducted with real data set of OWL-S Technical Chart (OWLS-TC). About 67% noise responses from syntactic search (N-Gram String Distance algorithm) are reduced in the proposed architecture. The results also illustrate the proposed architecture's capabilities of concept learning and so about significant improvement in service discovery after a few social concept contributions.