Semantic Infrastructure for a Ubiquitous Computing Environment By

This thesis investigates one of the fundamental problems for Ubiquitous Computing: managing metadata to enable resource discovery. This work presents a flexible and general model of metadata, and proposes to use ontologies as a formal language for the metadata of a Ubiquitous Computing Environment. Ontologies provide a common language for metadata among diverse and autonomous entities and spaces in the Ubiquitous Computing Environment. Description Logic is introduced as a formal language for metadata. Description Logic can be used to represent the concepts of a model, and the formal semantics can be used to maintain logical consistency. This thesis presents important locality principles: a Ubiquitous Computing Environment is a local space, which needs a dynamic “working set” from the hypothetical universe of possible devices, services, and entities. This concept is familiar from other contexts (e.g., memory management), but has not been recognized or used in ontology based systems. These principles provide a key insight that led to development key algorithms for managing ontologies in a Ubiquitous Computing Environment: an algorithm for composing two ontologies, and semantic queries on ontologies. The composition algorithm is the key to dynamically updating a local ontology to maintain a working set of concepts from a larger universe of ontologies. The composition algorithm exploits the formal semantics of the ontologies to maintain logical consistency when combining ontologies from several sources. The logical relations defined in the ontology enable semantic queries which can discover not just exact matches but logically related concepts. An improved algorithm for semantic matching is proposed, which extends and refines previous work from the literature, using the formal semantics of the ontology to define what concepts are similar to each other. A prototype Ontology Service was built using standards and software from the “Semantic Web”. The prototype implements algorithms for managing ontologies. The prototype was evaluated. In other work, the prototype Ontology Service was ported to Gaia.