Automatic generation of relational to ontology mapping correspondences

Abstract

This thesis presents Milan, an automatic relational-to-ontology system. Milan automatically generates mapping correspondences from a source relational database (RDB) and a target ontology. It addresses the relational-to-ontology challenges by resolving naming conflicts, structural and semantic heterogeneity. This enables high fidelity mapping correspondence generation for realistic databases that are de-normalised or utilize features of the relational data model that do not trivially map to RDF. Through a systematic review of the state-of-the-art in relational-to-ontology patterns and automatic relational-to-ontology mapping correspondence systems, the thesis sums up requirements automatic systems should address and also identifies the gaps present in the current state-of-the-art. The thesis leverages this analysis to design Milan, constituting three main processes and six sub-processes. The description of each process is done using UML process diagrams, algorithms and input/output templates. The thesis uses a popularly used benchmarking tool to evaluate Milan and other state-of-the-art systems. The evaluation experiment involves scenarios and tests, which not just describe the overall performance of a system but also quantitatively describes the various mapping challenges where the system fairs well and poorly.