Research devoted to developing knowledge-based tools for database design has demonstrated that it is possible to encode a great deal of process knowledge about database design in a (knowledge-based) computer program. However, experience with these tools shows that the contribution of an expert human designer extends beyond his or her knowledge of database design techniques. The paper discusses the application of an approach, called Naive Semantics (NS), to simulate the contributions made to a design based on the designer's general knowledge. Naive semantics involves the use of an extensible store of generally understood knowledge about the world. The paper describes the types of information that could be included in a Naive Semantics knowledge base and how that knowledge might be applied to increase the effectiveness of automated database design systems. Results of various implementations of ontologies based on Naive Semantics are discussed.

Although it is possible to encode a great deal of process knowledge about database design into a system, experience has shown that the contribution of a human designer extends beyond his or her knowledge of database design techniques. The next step in the evolution of automated database design tools is to incorporate knowledge and reasoning capabilities to support this higher level of participation. Doing so requires some understanding of what different terms mean. This paper presents an ontology that can be used as a surrogate for the meaning of words in a database design system to simulate the contributions that a designer would make based on his or her general knowledge. The ontology classifies a term into one or more categories such as person, abstract good or tradable document. It is comprised of a semantic network, a knowledge base containing information on the meaning of terms that have been classified, an expert system knowledge-acquisition component, and a distance measure for assessing the distance between the meanings of terms. The ontology was tested by different types of users on a variety of problems and was shown to be quite effective.