The Semantic Web is a vision that extends the present Web by adding formal semantics to its information. It opens up boundless possibilities by enabling software agents to reason about Semantic Web content and provide rational responses to unanticipated situations. Acceptance by people outside the technical community is critical to the success of the Semantic Web (Lassila et. al., 2001), however an obstacle to the successful deployment of the Semantic Web has been that one of its main building blocks, ontologies, are difficult to build (Hovy et. al., 2001). To address this problem, one needs to explore techniques that ease the process of ontology development while maintaining its critical qualities, like extendibility, adaptability and interoperability.
Reuse is one such technique. Reuse saves time and effort, while naturally lending itself to standardization. In some cases, a whole ontology can be built just from assembling other ontologies through reuse. In other cases, the reused ontologies must be extended, customized or adapted (Pinto et. al., 2000). Ontology engineering literature views reuse mostly as a manual activity that is dependent on the experience and skill of the ontologist(s). During an ontology development effort, an ontologist will usually browse a few ontology repositories and based on her experience with the repository and her understanding of the ontologies that are available to her, will select ontologies that can be used as is, or adapted with extensions. Although this is a simplistic view of a non-trivial task, this browse and select model can be characterized as the most common reuse effort that is being practiced in the field today. The emergence of the Semantic Web will see ontology libraries grow in size and extent, culminating in the Semantic Web itself becoming a global repository of ontologies. The current surf and select mode by expert ontologists will become less effective, and simple search agents alone are not the answer. I want to explore the possibility of using intelligent agents to address this crucial problem.
Although my research agenda needs refinement, I would like to work on developing agents that provide automated categorization and access to ontology repositories. This may provide significant improvement over existing Semantic Web ontology repositories that provide strictly hierarchical organization (Ding et. al. 2002)