Web Service Retrieval by Structured Models

Much of the information available on theWorldWideWeb cannot effectively be found by the help of search engines because the information is dynamically generated on a user’s request.This applies to online decision support services as well as Deep Web information. We present in this paper a retrieval system that uses a variant of structured modeling to describe such information services, and similarity of models for retrieval. The computational complexity of the similarity problem is discussed, and graph algorithms for retrieval on repositories of service descriptions are introduced. We show how bounds for combinatorial optimization problems can provide filter algorithms in a retrieval context. We report about an evaluation of the retrieval system in a classroom experiment and give computational results on a benchmark library.