Increasingly, ontologies are being developed and exposed on the Web to support a variety of applications, including biological knowledge sharing, enhanced search and discovery, and decision support. This proliferation of new Web knowledge sources is resulting in a growing need for integration and enrichment of these sources. Automated and semi-automated solutions to aligning ontologies are emerging that address this growing need with very promising results. However, nearly all approaches have focused on aligning ontologies-using relationships of similarity and equivalence and none have applied knowledge in upper ontologies. None to our knowledge have applied Support Vector Machine (SVM) technology. Only very recently, solutions for scalability of ontology alignment have begun to emerge.;The goal of this research is to advance the state of the art in automated ontology alignment by contributing in three main areas. First, we present new algorithms to extend the information can be derived in ontology alignment; specifically, new relationships between ontological components beyond similarity and equivalence. In particular, we present algorithms to align ontologies using subclass, superclass and relations contained within the original ontologies. We show how ontology alignment can be modeled in a Support Vector Machine and that use of SVMs enhances the ontology alignment process. Second, we contribute new evidence for ontology alignment. We show that the use of semantics in conjunction with upper ontologies and other linguistic resources can enhance the alignment process and specifically contribute to the discovery of new relationships cross-ontology. Finally, we investigate scalability issues in the area of processing, reasoning, and aligning large-scale ontologies. We present an alignment algorithm that bounds processing by selecting optimal subtrees to align and show that this significantly improves efficiency without major reduction in precision.
展开▼
