Mediation of XML Data Through Entity Relationship Models

This paper describes an approach for the querying of heterogeneous XML resources using an ontology-based mediator. Here an ontology is an Entity-Relationship schema defined independently of the schemas of the data sources. The sources are described to the mediator by means of mapping rules as in the Local-As-View approach to data integration. User queries are conjunctive queries formulated in terms of the ontology, and answers to these queries are obtained by rewriting them to XQuery expressions and evaluating these on the data sources. A formal semantics for queries is defined by interpreting XML sources into ER models. As there can be many such interpretations, a certain answer to a query is one which is true in all of them. We describe the rewriting algorithm and we show its completeness and correctness with respect to the given semantics. We also give an algorithm for producing a canonical model of the ontology and the interpreted data sources. It is shown that the certain answers can also be obtained by evaluating the query to just this one model.