The focus of information integration systems providing single access to data from distributed, diverse, and autonomous sources (including web-based sources) has changed from syntax and structure to semantics, allowing more meaningful integration of data. InfoQuilt1 goes one step further to support knowledge discovery by providing users with tools to analyze the data, understand the domains and relationships between them, and explore new potential relationships. It provides a framework to model the semantics of domains, complex semantic relationships between them, characteristics of available sources, and provides an interface to specify information requests that the system can “understand”. This thesis focuses on the use of knowledge about domains, their relationships, and sources to efficiently create practical execution plans for such semantic information requests.