Data semantics: data integration and the semantic web

Information systems today need to resolve the heterogeneity among data residing in multiple autonomous data sources. In particular, the use of the World Wide Web as a universal medium for exchanging information has radically transformed our vision about data access and manipulation. Following the perspective of the semantic web, we believe that the explicit presentation of data semantics will facilitate data interoperation in a variety of data manipulation tasks. Investigating and capturing the meaning of data is a core problem in all of applications in computer science. Especially, in database area the problem has been studied by many researchers. As a result, we have seen much progress on topics such as schema integration, schema matching, schema mapping, and generic model management in multidatabases, federated databases, and data integration over past two decades. The solutions that have been developed so far are a mixture of semi-automatic, heuristic-driven, and multi-layered solutions. Representing and reasoning about human knowledge in terms of symbolic structures is another long-standing problem. Consequently, many logical formalisms have been proposed bearing distinct expressivity and reasoning capabilities. A recent application of this work called semantic web aims at providing a machine-understandable Web by organizing data in terms of their semantics. It employs the approach of annotating data with formal ontologies. This approach demonstrates one aspect of the paradigm of ontology-based information representation and integration. We believe that investigating data semantics as to establish and maintain the correspondence between data/model and its intended subject matter could have long-term benefits as well as instant gains. Our future line of research is founded on a principle of semantic correspondence continuum. The continuum is grounded on domain ontologies.

Download PDF