Semantic matching across heterogeneous data sources

The role of semantic correspondences in semantic integration of data sources and to data integration across disparate databases are discussed. The growth of the Internet has increased the need for semantic interoperability across heterogeneous data sources. Semantic correspondences across heterogeneous data sources include schema-level correspondence and instance-level correspondence. Cluster analysis techniques are more suited to identify schema-level correspondence and classification techniques are more suited to detecting instance-level correspondences. Semantically related attributes tend to be highly correlated and can be identified through correlation analysis. Regression analysis can then be used to determine the actual relationship among correlated attributes. Corresponding records can be integrated into a single data set so that statistical analysis can be used to further analyze the relationships among attributes.