Editorial: Ontology Challenges: A Thumbnail Historical Perspective

Abstract

Rarely in the history of computer science has there been an idea that was so obviously right and that has failed for so long to deliver on its promise. In its simplest form, this idea can be phrased as follows: People use hierarchically organized knowledge about domain objects and relations as a major ingredient of their intelligent behavior. Thus, one might reason abductively, if we equip computers with this kind of knowledge and appropriate processing algorithms, we should get intelligent behavior. The earliest history of artificial intelligence concentrated on the processing algorithms. But, after many unexpected difficulties, research turned more and more toward the knowledge part of the knowledge plus reasoning algorithms framework. Yet, paradoxically, even though large amounts of knowledge have been amassed in some areas, notably medical informatics, and even though knowledge structures have been studied for over 35 years, the expected big breakthrough toward computational intelligence has not happened. This leaves us with the initially mentioned paradox: Why has an idea that is so obviously right not led to the expected results? In 1968, Quillian (Quillian 1968) published his famous semantic network paper. His network was essentially an electronic representation of a dictionary of real-world terms. Using that dictionary and spreading activation reasoning, Quillian’s program was able to generate simple natural language expressions of stored semantic relationships. Given the extremely limited hardware (in today’s view) that Quillian had available, his achievement was a major milestone that inspired research in semantic networks (Woods 1975; Lehmann 1992; Sowa 1991), cognitive psychology (Collins and Quillian 1969; Collins and Loftus 1975; Anderson 1983) and a field that was to become known as knowledge representation (Brachman and Schmolze 1985; Cercone and McCalla 1987; Sowa 2000).