UBC Theses and Dissertations
A knowledge level user interface using the entity-relationship model Chan, Hock Chuan
The relational database system has achieved widespread popularity; however, it is still very difficult for users, even those trained in the relational model, to formulate relational queries. The major cause of the difficulties is the fact that the user and the database system communicate using constructs that are not closely related to the user's world. This dissertation develops a new level of user-database interaction — the knowledge level (KL) interface — where the user and the database system exchange only knowledge of the domain. The data structure used in the database is fully hidden from the user. In this way, the query is very closely related to the user's world. Under the new KL approach, the database system is no longer seen as a store of data. It is set up as an agent to know the domain knowledge told to it by the user. The system will then provide the knowledge required by the user during retrievals. It will use elements of the entity-relationship model for communicating knowledge about the real world with the user. It is shown that the KL interface is in many ways better than the relational interface. Users of the KL interface need to know less and perform fewer data manipulation operations than users of the relational interface. The KL interface also achieves both physical and logical data independence, unlike the relational system which does not truly achieve logical data independence. This dissertation also proposes a new approach to understanding the meaning of completeness of a query language, breaking away from the traditional calculus-based measure of completeness. This new approach is then applied to the development of the knowledge level interface. The main contributions of this dissertation are the proposal and development of a knowledge level interface, the anaylsis to show that this interface is better than the relational interface, and the demonstration that such an interface is feasible even for large databases.