Light-weight semantics and Bayesian Classification: A hybrid technique for dynamic Web Service discovery

Abstract

Web Service discovery and ranking has been one of the key issues in Service Oriented Systems. Enormous efforts and research has been done towards semantic modeling of Web Services and a couple of semantic matchmaking and reasoning mechanisms have been developed to allow service consumers search for the required service providers dynamically. These approaches seem to be promising in theory, provided that exhaustive semantic descriptions of the services are available. However, in practice, this is not the case, as current Web Service standards provide quite limited information about services. Therefore, the process of discovery as well as ranking cannot always rely only on the extensive semantic descriptions to be available all the time. However, description of services using light-weight semantics (i.e., non-functional properties) is rather easier to have, and this could be used by classification and machine learning techniques to help in the classification of Web Services at real-time. In this paper, we present a hybrid approach towards enabling dynamic Web service discovery which is based on Bayesian Classification mechanism that classifies different available Web services, representing service providers, based on light-weight semantic descriptions.