Combining uncorrelated similarity measures for service discovery

In this paper we present an OWL-S service matchmaker that combines uncorrelated similarity measures for obtaining the services that match a given request. These similarity measures are obtained comparing four of the elements presented in the OWL-S Service Profile: name, description, and the collection of both inputs and outputs of a service and a request. For each of these elements a number of similarity measures can be applied and combined in several formulas in order to obtain a similarity value. Once these measures are calculated a neural network is trained to combine the uncorrelated similarity measures with the purpose of obtaining a degree of the suitability of a given service for a particular request. This matchmaker has been validated in the OWL-S TC v3 service library.

keywords: Web services, service matchmaker, similarity measures, neural network, OWL-S