| Summary: | There is pressing need for effectively integrating information from an ever increasing number of available sources both on the web and in other existing systems. A key difficulty of achieving this goal comes from the pervasive heterogeneities in all levels of information systems. Existing and emerging technologies, such as the Web, ODBC, XML, and Web Services, provide essential capabilities in resolving heterogeneities in the hardware and software platforms, but they do not address the semantic heterogeneity of the data itself. A robust solution to this problem needs to be adaptable, extensible, and scalable.
In this paper, we identify the deficiencies of traditional approaches that address this problem using hand-coded programs or require complete data standardization. The COntext INterchange (COIN) approach overcomes these deficiencies by declaratively representing data semantics and using a mediator to create the necessary conversion programs using a small number of conversion rules. The capabilities of COIN is demonstrated using an intelligence information integration example consisting of 150 data sources, where COIN can automatically generate the over 22,000 conversion programs needed to enable semantic integration using only six parametizable conversion rules. This paper makes a unique contribution by providing a systematic evaluation of COIN and other commonly practiced approaches.
|
|---|