This paper presents a novel approach to expedite a service discovery process. Since current service discovery approaches do not assume any preprocessing when service providers publish services into a service registry, service composition occurs at discovery time. Also, easily reusing a composed service for future discovery has not been considered. In this paper, instead of considering a service registry as a set of published services, we build a service knowledge base at publication time, which we call a service community. The service community consists of a set of service populations and their relationships in composition. A service population is a set of services in which the input and output parameters of all services are semantically equivalent. These pre-composed services will expedite the on demand service discovery process. Also a newly composed service is injected into the service community, and easily reused for future discoveries.