User-Database Interface: The Effect of Abstraction Levels on Query Performance
A common classification of data models is based on their abstraction levels: physical, logical and conceptual. The user-database interaction can be similarly classified. For the conceptual-level interaction, the user and the database exchange information on the user’s world, e.g., information of entities, relationships, and attributes. For the logical-level interaction, the user and the database communicate based on concepts in the database system, e.g., relations and join operations. We expect users to be familiar with concepts in their world but not the concepts in the database system. This is especially so for infrequent or naive database users. The conceptual level should therefore be easier because it is semantically closer to the user. This deduction was tested in an experiment using the entity-relationship (ER) model for the conceptual-level model and the relational model for the logical-level model. The results were affirmative. The users at the conceptual level had 38 percent higher accuracy and 16 percent higher confidence than users at the logical level. The conceptual-level users took 65 percent less time than the logical-level users, and it took 33 percent less time to train them. The differences were statistically significant with p < 0.003. The huge differences indicate that noticeable improvements can be made by switching from the relational model to the ER model. The experiment also provided valuable data on errors commonly made by users.
|Author||Hock Chuan Chan, Kwok Kee Wei, and Keng Leng Siau|
|Keywords||Data resource utilization, user-database interface, abstraction levels, conceptual level, logical level, relational model, entity-relationship model, query languages, SQL, experimental study, user performance|