Combining Scalability and Expressivity in the Automatic Composition of Semantic Web Services

Abstract

Automatic Web service composition (WSC) is a key component of flexible SOAs. We address WSC at the profile/capability level, where preconditions and effects of services are described in an ontology. In its most expressive formulation, WSC has two sources of complexity: (A) a combinatorial explosion of the services composition space, and (B) worst-case exponential reasoning is needed to determine whether the underlying ontology implies that a particular composition is a solution. Any WSC technology must hence choose a trade-off between scalability and expressivity. We devise new methods for finding better trade-offs. We address (A) by techniques for the automatic generation of heuristic functions. We address (B) by approximate reasoning techniques for the fully expressive case, and by identifying a sub-class where the required reasoning is tractable. We show empirically that our approach scales gracefully to large pools of pre-discovered services, in several test cases.