Domain Speci c Ontologies for Semantic InformationBrokering on the Global Information

Recent emerging technologies such as internetworking and the World Wide Web (WWW) have signiicantly expanded the types, availability, and volume of data accessible to an information management system. In this new environment it is imperative to view an information source at the level of its relevant semantic concepts. We propose that these semantic concepts be chosen from pre-existing domain speciic ontologies. Domain speciic ontologies are used as tools/mechanisms for specifying the ontological commitments or agreements between information users and providers on the information infrastructure. We use domain speciic ontologies to tackle the information explosion by the: (a) Re-use and organization of knowledge in pre-existing real world ontologies, achieved by mapping semantic concepts in the ontologies to data structures in the underlying repositories; and (b) Knowledge integration and development of mechanisms to translate information requests across ontologies. We thus provide support for multiple domain speciic ontologies as alternate world views on the vast amounts of data. Semantic information brokering is implemented by brokering across domain ontologies based on interontology relationships such as synonyms, hyponyms and hypernyms deened between terms in diierent ontologies. Information requests are rewritten using these relationships to obtain translations across ontologies. These ideas have been implemented in the OBSERVER system, the algorithms for which are discussed in this paper.