Welcome to the Upper Cyc® Ontology

E-Mail Comments to: doc@cyc.com
Last update: August 12, 1997
This web page is a beta-version release only, subject to change in the near future.
Copyright© 1996, 1997, 1998, 1999 Cycorp All rights reserved. See Terms of Usage.
Cyc® is a Registered Trademark of Cycorp, Inc.

[NEXT PAGE]

Cycorp welcomes you to its first major public release: approximately 3,000 terms capturing the most general concepts of human consensus reality. We refer to this as the "upper Cyc® ontology." The full Cyc® knowledge base (KB) includes a vast structure of more specific concepts descending below this upper level. Over the past dozen years, we have also entered into Cyc® literally millions of logical axioms -- rules and other assertions -- which specify constraints on the individual objects and classes found in the real world. Further specializations have been developed for our customers, especially in recent years, driven by their application needs. Thus, only a few percent of Cyc's terms appears in this release (and with only a small portion of their content). Later on, on the next page, we give a more detailed list of the parts of Cyc® omitted from this public release.

But we believe these 3,000 terms to be of great importance, interest, and use. Why? Because it is the core of an ontology that satisfies these two important criteria:

It is "universal:" Every concept one can imagine can be correctly linked into the Upper Cyc® Ontology in appropriate places, no matter how general or specific, no matter how arcane or prosaic, no matter what the context (nationality, age, native language, epoch, childhood experiences, current goals, etc.) of the imaginer.

It is "articulate:" The distinctions which are made in the ontology are both necessary and sufficient for most purposes: By "necessary" we mean that the distinctions are all worth making. There is both theoretical and pragmatic justification for every collection, for every predicate and function, for every individual. By "sufficient" we mean that enough distinctions have been made to enable and support knowledge sharing, natural language disambiguation, data base cleaning and integration, and other applications discussed below.

We do not make this claim lightly. Our conviction is based on person-centuries of effort expended over the past dozen years in constructing the Cyc® KB and applications using it. Our selection of Cyc's upper-level concepts has been tested in the crucible of extensive logical deduction on tens of millions of examples. In that process, many categories and functions that initially seemed useful proved not to be (like #$CompletelyTangibleObject and #$transfersThrough), while other "uglier" categories and functions (like #$SiblingDisjointCollection and #$SkolemFunction) emerged as surprisingly powerful and cost-effective and long-lived. Thousands of our most important discoveries are here and now made available to others.

The Upper Cyc® Ontology provides a sufficient common grounding for applications of the following types:

Natural language understanding and generation [footnote];
Semantic database integration, consistency-checking, and mining
Semantic information retrieval
Ontology-constrained simulation
Building and utilizing user models
Knowledge sharing by groups working independently

and so on. Our special intent is to address readers who are developing such applications. We do not aim only at those who will be using the complete Cyc® system. Even systems that are differently motivated and arranged can benefit in practical ways from an accurate logical mapping of their ontologies and term hierarchies to Cyc's upper ontology. The materials in Release 2.1 should provide the basis for readers to determine where and how each new application-specific concept would fit into the general Cyc® ontology. That alone should suffice to give you some boost in building and sharing the aforementioned sorts of applications. Of course we hope and expect that many of you will then see the value in having access to the other 99% of Cyc®, and we welcome such inquiries into Cyc's availability.

Every new concept will have some related concepts somewhere in the upper Cyc® ontology. One does not just add a new term to the ontology, one states assertions about it that serve to "situate" that new concept properly in the existing ontology. This does not, however, mean fully defining the new term; indeed, most of the terms worth having will never be precisely defined (e.g., "white collar worker.")

It is assumed that other systems will share the same upper ontology with Cyc® but diverge somewhat at lower levels, where more specialized terms are introduced and organized. That is neither a fundamental problem nor an obstacle to future integration of ontologies. Cf. our comments above, about Cyc® being universal and articulate. When concepts in other systems have been correctly mapped to their related concepts in the upper Cyc® ontology, differences should be confined to small, subject-related areas in which there may be some divergence in definitions.

And, most importantly, Cyc® permits (and even expects) its knowledge to be structured into lattices of reified, formalized, "first class" contexts. Such Cyc® "microtheories" can represent different points of view, levels of granularity, cultural differences, age differences, time periods, corporate cultures, etc.

For example, ten idiosyncratic and mutually inconsistent "models" for the US economy could be accommodated within Cyc® as ten separate microtheories, whose local collections and relations would all link upwards to appropriate generalizations in the upper Cyc® ontology. In each of those microtheories, terms would have different (partial) definitions, different rules which were stated about them, different assumptions that were presumed to hold true, etc.

We provide several tools to assist readers in becoming familiar with the upper Cyc® ontology, as the next page explains:

Concise English descriptions for all 3,000 collections, individuals, predicates, relations, and functions;
Hypertext links to every other relevant item in the release, wherever it appears (e.g., inside a comment, as a superset of some concept, or as an argument type restriction). The links enable both direct browsing of some of the taxonomic (subset/superset and elements/element-of) trees and navigation among the references within comments;
A topical listing of the upper ontology divided into subject-areas, to facilitate systematic study;
Some atypically detailed examples of subclasses and instances, here and there in selected subject areas, to give a clearer picture of intended uses, meaning, and coverage.

Please let us know if you are trying to include some concept that doesn't seem to be readily accommodated in the upper Cyc® ontology, or if you find one of the existing terms' comments unintelligible. For more information about Cycorp, and the Cyc® technology, click here.

We're excited about publishing the Upper Cyc® Ontology, and hope that you find it useful.

--- the Cycorp technical staff

We have prepared this document as a set of HTML pages. The first page, after this cover letter, is "The Cyc® Ontology Guide: Introduction". We will shortly be making even more of the Upper Cyc® Ontology available, publicly, along with downloadable cgi code for browsing through the hierarchies on your local machine. For now, such things are available to Cycorp customers and strategic partners. We wish to thank IBM in particular for sponsoring the preparation of Release 1.0 and 2.0.

Go to Top