Approximate ontology merging for the semantic Web

Abstract

The semantic Web is the next step in the Internet's evolution. In order to facilitate autonomous information retrieval, it is necessary to create a standard framework for the semantic Web. The existing Web contains a considerable amount of data. It is an alluring target for data mining. Most of the data stored on Web pages is weakly structured. Mining or otherwise accessing the data in context from multiple sites is difficult; particularly when others construct the Web pages and their internal structure is not clearly understood. A standard way of viewing and organizing data into an ontology is essential for sharing data and interoperability between Web sites. A semantic Web standard has evolved through extensive iteration and was approved in late 2003. Fully autonomous or semi-autonomous discovery and development of ontologies is the only feasible way to transition to the semantic Web. One enabling strategy is to merge existing, vetted ontologies into a single ontology. The pre-merged ontologies would likely be similar in some respects and distinct in others. This paper discusses how to possibly merge multiple ontologies into a single viable ontology. It seeks an effective way to merge ontologies without losing any information (and possibly gaining some information in the process). When comparing ontologies, a similarity measure would be necessary. The measure would be used to measure the similarity two existing ontologies. This similarity measure is necessarily imprecise.