General Competition (October 1997)

A National Knowledge Infrastructure

Sponsor: Cycorp, Inc.

• Project Performance Period: 10/1/1997 - 9/30/2000
• Total project (est.): $3,069,367.00
• Requested ATP funds: $1,640,000.00

One of the steepest barriers to realizing the full potential of computers to enrich and simplify our lives is that they don't know the millions of things that we assume "everyone" knows, such as that most people sleep at night, or that infants can't drive cars, or that you can't be in Dallas and Chicago at the same time, or that you can't be as old as your mother. However, it is possible to represent such knowledge in ways that computers can use it. Cycorp will work on this task and will develop software tools that conduct clarifying dialogues to zero in on what the user intends. Together, this tool and the underlying knowledge base should enable the average person to express their requests in plain English and to better have those requests served. For instance, online text searching programs often return a huge number of irrelevant matches with an unknown, unseen, but equally large number of missed opportunities that are never located. Even a small degree of understanding of the query and the material being searched through can eliminate the vast bulk of those false positives and false negatives. Within the nation's evolving information infrastructure, Cycorp proposes to build the foundation elements of a "widely held information layer," or WHI: a large reference ontology along with interface and inference tools that allow people to interact with this repository of knowledge, comment on and extend it, and have it deduce new conclusions for them from it. The company will use Cyc--the large and growing common sense knowledge base begun in 1984--and its CycL representation language as the starting points. Cyc is already the largest knowledge base in the world, but it must be expanded even more to attract and justify the efforts of other organizations and individual software developers. In part, this will entail integrating other ontologies, most of which are highly specialized, into the WHI. Cycorp's goal is to provide a catalyst and "jump start" an assortment of bottom-up, third-party efforts to bring knowledge-based applications to market. The technical challenges include developing multiple means of expressing the same idea without sacrificing efficiency, accommodating divergent viewpoints without getting tangled in contradictions, as well as defining the hundreds of thousands of basic terms and millions of basic rules that make up an adequate core on top of which others can build. In the near term, tools developed by Cycorp with ATP support can be of use to almost any firm struggling to integrate multiple databases, a market that is forecast to grow to $21 billion by the year 1999.

For project information:

Douglas B. Lenat, (512) 342-4001
lenat@cyc.com

ATP Project Manager

Barbara Cuthill, (301) 975-3273
barbara.cuthill@nist.gov