Newsgroups: alt.hypertext,comp.ai
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!pipex!uunet!world!Eastgate
From: Eastgate@world.std.com (Mark Bernstein)
Subject: Re: Hypertext and "Static AI"
Message-ID: <D84o0D.Mw0@world.std.com>
Organization: The World Public Access UNIX, Brookline, MA
References: <3odiks$2q0@Venus.mcs.com>
Date: Fri, 5 May 1995 23:02:37 GMT
Lines: 52

As Borger observes, the interface between hypertext and knowledge
representation -- the creation of data structures that represent
ideas in rich and meaningful ways -- is a fascinating and important
area. Some very interesting work has been done on this topic over
the years, and some alt.hypertext readers might find this research
useful as well as interesting.

A good starting point is Lenat's CYC project, a ten-year effort to
create an extensive knowledge representation of common-sense 
knowledge -- the knowledge that everyone assumes everyone else
is likely to know. At the 1988 AAAI Workshop on AI and Hypertext,
Lenat gave a very interesting discussion of the relationship 
between hypertext and AI, suggesting that hypertext was a simple
but incomplete solution to the difficult and important problems that
CYC and similar project address. (Ted Nelson, in his talk, espoused
the One True System; Lenat termed CYC's goal "one true representation").
CYC is discussed extensively in 

  Lenat, D. B.; Guha, R.V.; Building Large Knowledge-Based Systems: 
  Representation and Inference in the CYC Project, 
  Reading, Addison-Wesley, 1990.

A hypertextual front-end to CYC is presented by Michael Travers 
(Hypertext 89);  see also Clitherow's Hypertext 89 paper for related
work. Bieber's Hypertext 91 paper discussed an interesting system
that uses domain knowledge (for material procurement) without
the deep analysis of CYC.

The extensive node and link typing in NoteCards (Halasz and Trigg,
HT 87 and CHI 87) can be viewed as an effort to merge knowledge
representation and hypertext. My earlier comments on the fate of 
these link types were clearly too harsh; research interest in the idea
waned in the early 90's, but the underlying ideas remain influential
and may be experiencing a revival. See, for example, Jordan and
Russell's HT89 paper on IDE, an extension to NoteCards that adds
composite types and templates, as well as Streitz et al.'s SEPIA.

Nanard and Nanard's MacWeb (HT91, HT93, ECHT 94, no relation
to the web browser of the same name) features facilities that allow
writers to declare structure *and* also facilities that induce structural
descriptions from examples. For example, a writer can tell MacWeb,
"This is a catalog page. This is another catalog page. And this is one,
too."; MacWeb will build a description of the abstract notion "catalog
page" which can be edited and instantiated automatically. 



-- 
Mark Bernstein
Eastgate Systems, Inc.   134 Main Street   Watertown MA 02172 USA
voice: (800) 562-1638 in USA   +1(617) 924-9044
Eastgate@world.std.com    Compuserve: 76146,262    AppleLink:Eastgate 
