With the growth of business demands, it becomes crucial to compose existing services in order to create new value added applications. Some languages like BPEL, WSCI enable creating complex composite services but they only allow static, manual composition. Basing on semantic technologies, many researchers have proposed (semi-)automatic methods for service composition, which often rely on a close world assumption and are not scalable for complex systems. In essence, these approaches lack a means for users expressing their complex business demands in a form that facilitates systems automatically generating a composite service. In this paper, we introduce our approach to provide an ontology/language based on the OWL, called OWL-T, for formally and semantically defining task templates that capture business demands at a high-level abstraction. Therefore, with a complex business demand expressed in terms of structured tasks, users need not to care about technical details of composition languages and an automatic composition method can be used to transform task templates into executable processes. In addition, based on semantic technologies, the OWL-T ontology also enhances the interoperability for better sharing and reuse of task templates in various domains of discourse.