Applying Principle of Locality to a Knowledge Base of Schema Mapping

Large amounts of information are posted on the web every day by thousands of enterprises, organizations and individuals. There is a daily exchange of data between companies, who also proceed to an extraction of data from web databases and documents in order to integrate them into databases. Since XML has become as the de facto standard in this area, the data contained in XML files are loaded into the database, usually relational, to be processed or analysed. In order to enable different systems to communicate with each other, it is necessary to match the data and create a mapping between the source schema and the target schema. Yet, in many companies, this mapping is manually built. Loading XML data, that were previously submitted to a manually constructed mapping, in databases, is subject to a prior itself automated schema matching. We propose in this paper a solution for loading (semi-)automatically XML data into relational databases. The solution we propose is to use previous mappings. These mappings are stored in what we call a Mapping Knowledge Base. We study the evolution of the volume of the Mapping Knowledge Base and we propose a strategy to use.