Three principles of representation for semantic networks

Abstract

Semantic networks are so intuitive and easy to use that they are often employed without much thought as to the phenomenon of semantic nets themselves. Since they are becoming more and more a tool of artificial intelligence and now database technology, it is appropriate to focus on the principles of semantic nets. Such focus finds a harmonious and consistent base which can increase the semantic quality and usefulness of such nets. Three rules of representation are presented which achieve greater conceptual simplicity for users, simplifications in semantic net implementations and maintenance, and greater consistency across semantic net applications. These rules, applied to elements of the net itself, reveal how fundamental structures should be organized, and show that the common labeled-edge semantic net can be derived from a more primitive structure involving only nodes and membership relationships (and special nodes which represent names). Also, the correlation between binary and n -ary relations is presented.