XML storage in relational databases: An approach combining description logic and statistics

We propose in this paper to jointly use the description logic (DL) and probabilistic approaches to store documents in relational databases. The description logic is used to generate a first database schema by reasoning capabilities over the conceptual part of the documents. The resulting schema is normalized. Indeed, we add to the DL knowledge base, which represents the documents, some ontological assertions specified in the DL formalism. Then probabilistic calculus is used to optimise the generated database schema by computing some statistical measurements over the extensional part of the documents.