The advent of Service Oriented Architecture makes services the most flexible, platform-independent choice for exposing and integrating business functionality across a network. However, the lack of service discovery mechanisms, that go beyond simple keyword search to enable automated late-binding of service requesters to providers, remains a major problem. Realistic late-binding involves matchmaking of client requests with service descriptions, based on frequently changing fine-grained client requests, and dynamically changing service functionality. The intricacies of service functionality cannot be specified by generic static descriptions since the functionality offered by the service may depend on the client at hand, their specific request and the service?s current capabilities. In this paper, we propose a semantic framework supporting dynamic data fetching from services during the discovery phase on service instance level entailing a communication overhead which we aim to minimize.