DIDSBS applies AI techniques to improve access to heterogeneous, distributed and semistructured information sources as they are presented in the WWW or organization-wide intranets. Our system uses ontology to access sets of distributed XML documents on a conceptual level and stores them in the relational database. We integrate conceptual modeling, inference mechanisms with XML. We present the mapping algorithms, which including mapping between ontology and DTD and mapping between DTD and database schema, and then implement a close correlation between XML documents and database schema on semantic and structure. Thus, we are able to supplement syntactical access to XML documents by conceptual supports.