Scalable Web Service Composition with Partial Matches

We will investigate scalable algorithms for automated Semantic Web Service Composition (WSC). Our notion of WSC is very general: it allows the generation of new constants by Web service outputs; the composition semantics includes powerful background ontologies; and we use the most general notion of matching, partial matches, where several services can cooperate each covering only a part of a requirement. Herein, we define a first formalism. We show that automatic composition is very hard: even testing solutions is Π p -complete. We identify a special case that covers many relevant WSC scenarios, and where solution testing is only coNP-complete. While coNP is still hard, in the area of planning under uncertainty, scalable tools have been developed that deal with the same complexity. In our next step, we will adapt the techniques underlying one of these tools to develop a scalable tool for WSC. In the long term, we will investigate richer formalisms, and accordingly adapt our algorithms.