) is a crucial phase in Data Warehouse (
) design life-cycle that copes with many issues: data provenance, data heterogeneity, process automation, data refreshment, execution time, etc. Ontologies and Semantic Web technologies have been largely used in the
phase. Ontologies are a buzzword used by many research communities such as: Databases, Artificial Intelligence (AI), Natural Language Processing (NLP), where each community has its type of ontologies: conceptual canonical ontologies (for databases), conceptual non-canonical ontologies (for AI), and linguistic ontologies (for NLP). In
approaches, these three types of ontologies are considered. However, these studies do not consider the types of the used ontologies which usually affect the quality of the managed data. We propose in this paper a semantic
approach which considers both canonical and non-canonical layers. To evaluate the effectiveness of our approach, experiments are conducted using Oracle semantic databases referencing LUBM benchmark ontology.