Leveraging Web Services Discovery with Customizable Hybrid Matching

Improving web service discovery constitutes a vital step for making the Service Oriented Computing (SOC) vision of dynamic service selection, composition and deployment, a reality. Matching allows for comparing service requests of users with descriptions of available service implementations, and sits at the heart of the service discovery process. During the last years, several matching algorithms for comparing user requests to service interfaces have been suggested, but unfortunately there are no consistent comparative experimental evaluations of existing service discovery methods. This paper firstly reviews several state-of-the-art approaches to matching using syntactic, semantic and structural information from service interface descriptions. Secondly, it evaluates the efficacy of several key similarity metrics that underpin these approaches, using a uniform corpus of web services. Thirdly, this paper develops and experiments with a novel style of matching that allows for blending various existing matching approaches and makes them configurable to cater service discovery given domain-specific constraints and requirements.