Discovering conceptual structure of relations

This paper presents an overview of a system to transform relational schemes into a conceptual structure which is represented in frames, and a mechanism for discovering conceptual structure of a currently existing database, named TRICS. In order to represent the conceptual structure, it is necessary to make conceptual objects and relationships between them clear. They are obtained by analyzing the relations, and by inferring the functional and inclusion dependencies inductively. The functional dependencies are used for making conceptual objects clear, as well as the inclusion dependency relationships between them. The system treats three types of relationships, they are: is-a, part-of, and association. A hierarchical data structure of frames is constructed based on an is-a, and frames are related with part-of and association. In this way, the system extracts structural representation from relations. As a result, the transparency of relations and its understanding are improved