A Methodology to Prepare Real-World and Large Databases to Ontology Learning

Several approaches have been proposed for the generation of application ontologies from relational databases. Most of these approaches, propose a fully automatic process based on unrealistic assumption, where the input database is well designed, up-to the third normal form (3NF). Real-World databases may contain irrelevant, missing or erroneous information to the ontology learning process. Preparing databases before ontology learning is quite rare. We propose in this paper a methodology for Database Preparation (DBP), composed of three sub-processes: the extraction of a Business Database (BDB), the cleaning of the BDB, and the enrichment of the cleaned BDB. A proof-theoretical case study shows that the proposed methodology is feasible and useful.