Semantic technology offers a promising tool to address the problem of today's information overload by interpreting the meaning of content and data for their automated processing within computer systems. In the Semantic Web, ontologies provide the vocabulary needed for such a meaningful and machine-understandable interpretation of data, and they are the basis for semantic annotation of web resources. As artifacts of knowledge representation, ontologies allow to reason about semantically described resources, and one way of reasoning about semantic descriptions is their matchmaking in terms of resource compatibility. A particular application of matchmaking in the Semantic Web is the discovery of Web Services annotated by descriptions of service functionality, where inference mechanisms are used to judge whether a service offer is compatible with a service request. The reasoning machinery provided by current Semantic Web language standards like OWL, however, lacks various forms of common-sense behaviour due to the monotonicity of the underlying description logic formalism, while nonmonotonic extensions to description logics allow for such common-sense reasoning.

In this thesis, we investigate several nonmonotonic extensions to description logics (DLs), namely autoepistemic DLs, circumscriptive DLs and terminological default rules, all of which extend standard DL inference mechanisms by forms of closed-world and default reasoning associated to common-sense features. We establish a matchmaking framework for semantic resource descriptions formulated in the DL formalism that uses various DL inferences to judge about resource compatibility. We put special emphasis on mapping the technical formalities of model-theoretic semantics of DLs to more intuitive notions that abstract from the details of logic for the framework's easier adoption in applications. Moreover, we incorporate the common-sense features realised by nonmonotonic DLs into the matchmaking mechanism to improve matching behaviour in situations where semantic resource descriptions are incomplete. Based on this framework, we apply the technique of matchmaking to the problem of service discovery in the Semantic Web with services annotated by OWL descriptions formulated in terms of domain ontologies. We evaluate the thus obtained service discovery mechanism by testing its applicability and usefulness in a concrete application scenario taken from a project case study in the logistics domain.

The particular contributions of this thesis span the fields of nonmonotonic reasoning with description logics in Artificial Intelligence, matchmaking of ontology-based descriptions and Semantic Web Service discovery. We introduce a novel tableaux calculus for reasoning in circumscriptive DLs, and we demonstrate how the various nonmonotonic extensions to description logics can be used to realise common-sense features and local closed-world reasoning in a Semantic Web setting in general. We characterise the use of various DL inferences for solving the matchmaking problem and show how different realisations of local closed-world reasoning help to overcome problems that arise due to the open-world semantics of OWL and classical DLs. We also provide initial methodological guidance to the modelling of semantic resource descriptions in the context of matchmaking. Furthermore, we position our semantic matchmaking technique as an adequate tool for supporting Semantic Web Service discovery and demonstrate its applicability in a concrete use case scenario.

In essence, we extend classical description logics by nonmonotonic reasoning capabilities, incorporate the resulting features of common-sense inference into semantic matchmaking mechanisms and apply this technique to the problem of Semantic Web Service discovery.