Knowledge-based system for relational normalization of GDBMS conceptual schemas

The knowledge-based system described attempts to entirely automatize the normalization process of a real-world database schema and to produce an equivalent, 'better' normalized scheme within the framework of a multilayered GDBMS (generalized database management system) architecture. The implemented system prototype has demonstrated the feasibility of the approach for small databases with simple schema, i.e. schemas with no high-level constructs. For more complex schemas, such as extended E/R schemas with generalization and aggregation high-level constructs, the project made apparent the need for the user to provide more domain-specific semantic knowledge beyond relation attributes, and functional and multivalued dependencies. The technical difficulties related to the interaction between high-level abstraction and normalization, the projection of global functional dependencies, and the management of the combinatorial explosion of decomposed relations have been analyzed.<>