Re: Doug Lenat's THE BIG PICTURE

• Mail folder: SRKB Mail
• Next message: Daniel D Suthers: "SGML & Ontologies (etc.) "
• Previous message: Fritz Lehmann: "Re: All agreed, but what's out there now?"
• In-reply-to: Sam Hunting: "Re: Doug Lenat's THE BIG PICTURE"

Date: Fri, 9 Jun 1995 09:04:58 +0500
From: sowa@west.poly.edu (John F. Sowa)
Message-id: <9506091304.AA00405@west>
To: sgmlsh@world.std.com
Subject: Re: Doug Lenat's THE BIG PICTURE
Cc: cg@cs.umn.edu, doug@csi.uottawa.ca, fritz@rodin.wustl.edu, hovy@isi.edu,
        lenat@cyc.com, srkb@cs.umbc.edu
Content-Length: 1374
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk

Sam Hunting writes,

>Isn't that one of the things that SGML (Standard Generalized Markup 
>Language, ISO 8879) would be useful for? That is, putting content into 
>useful, machine-readable, common forms usable by sophisticated 
>application software.

SGML is widely used for applications that process natural language
text.  It is a major improvement over the old formatting languages and
an even bigger improvement over the so-called WYSIWYG word processors.
The major limitation of WYSIWYG is that What-You-See-Is-All-You-Get.

SGML is important for identifying chapters and sections, keeping track
of the definition of a word from its pronuciation, and incorporating
various kinds of tags for many different purposes.  That is certainly
important, but it is very far from what Doug Lenat, Ed Feigenbaum, or
most AI researchers need.

Besides identifying the definition part of a dictionary entry,
we need some way of representing what it means, generally at a much
more detailed level than is needed for telling people what it means.
That is where AI knowledge representation languages come in.  They
address the very detailed semantics of the knowledge, not just the
major components of a document.

SGML and the AI knowledge representation languages are complementary.
They can and should be used together, but neither one is a replacement
for the other.

John Sowa