Mark James Carman and Craig A. Knoblock
We introduce a framework for learning Local-as-View (LAV) source definitions for the operations provided by a Web Service. These definitions can then be used to compose the operations of the Web Service into query execution plans. In order to learn the definition for a new service, we actively invoke the service and compare its output with that of other known services. We combine Inductive Logic Programming (ILP) and Query Reformulation techniques in order to generate and test plausible source definitions for the operations of the service. We test our framework on a real Web Service implementation.