Measuring the Quality of an Integrated Schema

Schema integration is a central task for data integration. Over the years, many tools have been developed to discover correspondences between schemas elements. Some of them produce an integrated schema. However, the schema matching community lacks some metrics which evaluate the quality of an integrated schema. Two measures have been proposed, completeness and minimality. In this paper, we extend these metrics for an expert integrated schema. Then, we complete them by another metric that evaluates the structurality of an integrated schema. These three metrics are finally aggregated to evaluate the proximity between two schemas. These metrics have been implemented as part of a benchmark for evaluating schema matching tools. We finally report experiments results using these metrics over 8 datasets with the most popular schema matching tools which build integrated schemas, namely COMA++ and Similarity Flooding.