Automatic Mapping Generation and Adaptation for XML Data Sources. (Génération et adaptation automatiques de mappings pour des sources de données XML)

The integration of information originating from multiple heterogeneous data sources is required by many modern information systems. In this context, the applications’ needs are described by a target schema and the way instances of the target schema are derived from the data sources is expressed through mappings. A mapping describes the transformation and the integration of the source instances to conform to the target schema. In this thesis, we address the problem of mapping generation for a XML target schema from multiple XML data sources and the automatic adaptation of these mappings when the target schema or the source schemas evolve. We propose an automatic generation approach that first decomposes the target schema into subtrees, then defines mappings, called partial mappings, for each of these subtrees, and finally combines these partial mappings to generate the mappings for the whole target schema. Applying our approach creates a set of alternative mappings: all for the given target from the sources. If the information required in the target schema cannot be provided by the sources, then no mapping can be generated. In this case, we propose to relax some constraints of the target schema such that mappings can be generated. We also propose a mapping adaptation approach to keep existing mappings current if some changes occur either in the target schema or in one of the source schemas. We have developed a prototype implementation of a tool to support the generation process. te l-0 03 24 42 9, v er si on 1 25 S ep 2 00 8