Capture database semantics by rule induction

To capture database characteristics, a knowledge-based entity-relationship (KER) model is proposed to extend the basic ER model by P.P.S. Chen (see ACM Trans. Database Syst., vol.1, no.1 (1976)) to provide knowledge specification capability. The knowledge specification capability allows database characteristics to be specified and maintained with each object definition. In the KER model, each entity or relationship has its specific characteristics. These characteristics can be classified into intraobject knowledge and interobject knowledge. Intraobject knowledge specifies how an object instance belongs to an entity type, and interobject knowledge describes how objects are correlated with each other when they are bounded by the same relationship. Instances of the database objects have to follow these rules since each database state is an instance of the application. Therefore, semantic knowledge can be induced from the database instances by machine learning based on the schema specified in the KER model. A knowledge acquisition methodology that is based upon the KER Model and machine learning techniques is developed to induce the database characteristics knowledge from the database