Combining schema and instance information for integrating heterogeneous data sources

Abstract

Determining the correspondences among heterogeneous data sources, which is critical to integration of the data sources, is a complex and resource-consuming task that demands automated support. We propose an iterative procedure for detecting both schema-level and instance-level correspondences from heterogeneous data sources. Cluster analysis techniques are used first to identify similar schema elements (i.e., relations and attributes). Based on the identified schema-level correspondences, classification techniques are used to identify matching tuples. Statistical analysis techniques are then applied to a preliminary integrated data set to evaluate the relationships among schema elements more accurately. Improvement in schema-level correspondences triggers another iteration of an iterative procedure. We have performed empirical evaluation using real-world heterogeneous data sources and report in this paper some promising results (i.e., incremental improvement in identified correspondences) that demonstrate the utility of the proposed iterative procedure.