Subsumption for Complex Object Data Models

We adopt a formalism, similar to terminological logic languages developed in AI knowledge representation systems, to express the semantics of complex objects data models. Two main extensions are proposed with respect to previous proposed models: the conjunction operator, which permits the expression multiple inheritance between types (classes) as a semantic property and the introduction in the schema of derived (classes), similar to views. These extensions, together with the adoption of suitable semantics able for dealing with cyclic descriptions, allow for the automatic placement of classes in a specialization hierarchy. Mapping schemata to nondeterministic finite automata we face and solve interesting problems like detection of emptiness of a classextension and computation of a specialization ordering for the greatest, least and descriptive semantics. As queries can be expressed as derived classes these results also apply to intentional query answering and query validation.