ON THE EPISTEMOLOGICAL STATUS OF SEMANTIC NETWORKS**Prepared in part at Bolt Beranek and Newman Inc. under contracts sponsored by the Defense Advance Research Projects Agency and the Office of Naval Research. The views and conclusions stated are those of the author and should not be interpreted as necessarily representing the official policies, either express or implied, of the Defense Advanced Research Projects Agency or the U.S. Government.

This chapter examines in detail the history of a set of network-structured formalisms for knowledge representation—the so-called “semantic networks,” which were introduced around 1966 as a representation for the concepts underlying English words and were designed expressly to treat concepts as formal representational objects.

Abstract

This chapter examines in detail the history of a set of network-structured formalisms for knowledge representation—the so-called “semantic networks.” Semantic nets were introduced around 1966 as a representation for the concepts underlying English words, and since then have become an increasingly popular type of language for representing concepts of a widely varying sort. While these nets have for the most part retained their basic associative nature, their primitive representational elements have differed significantly from one project to the next. These differences in underlying primitives are symptomatic of deeper philosophical disparities, and I discuss a set of five significantly different “levels” at which networks can be understood. One of these levels, the “epistemological,” or “knowledge-structuring,” level, has played an important implicit part in all previous notations, and is here made explicit in a way that allows a new type of network formalism to be specified. This new type of formalism accounts precisely for operations like individuation of description, internal concept structure in terms of roles and interrelations between them, and structured inheritance. In the final section, I present a brief sketch of an example of a particular type of formalism (“Structured Inheritance Networks”) that was designed expressly to treat concepts as formal representational objects. This language, currently under development, is called KLONE, and it allows the explicit expression of epistemological level relationships as network links.