Generic approach to knowledge management from databases

Abstract

Digital intelligence1 relies, among other things, on the involvement of software agents in the interpretation and reasoning on the information. This requires the immersion of meaning in information storage systems. Software agents can qualitatively treat information only if they can understand it, which is possible due to the cooperation of the ontology as source of knowledge and the data sources as an informational heap. In this paper, we present an automatic approach that allows to create an ontology from a classical data source (Relational, Object Oriented or Semi-Structured), and to interrogate the ontology thus created without resorting to creating instances: To create the ontology, the approach is based basically on meta-model, model and transformation concepts. Indeed, the different stages are translated by the transformation of an input model for the generation of an output model, this generation is directed by meta-models that we proposed. We present also a system of querying ontology without having to populate it with instances from data source. This system provides an intermediate level of abstraction between the ontological model and the data source schemas, this level can generate partially and temporarily the data in XML format. The system also provides a SPARQL-XQUERY mapping which rewrites any SPARQL query at XQUERY query in order to be executed on the generated data.