John Sowa asks:
> Cyc has done several person-centuries of work, and the ability to
> augment their knowledge base by reading the same kinds of books
> that people read was always one of their goals. They have had
> many top-notch linguists working with them. Yet they haven't
> achieved that. What are you proposing that they haven't done?
>
(1) To build a **completely open** system that has one common foundation
ontology structured as a Conceptual Defining Vocabulary and shown to be able
to create the logical specifications of the meanings of a term from an
English-language definition using the English defining vocabulary;
(2) to have that common foundation ontology open to inclusion of any
logically consistent representation from any source, with mappings or
translations where needed;
(3) to associate with that a language understanding system that is
word-oriented and so highly modularized that one can build and include a
word expert for a single word, replace the word expert for that word, and
test that new system against the same set of tests (possibly with additional
tests proposed as meaningful) to determine whether the new word expert is an
improvement on the older one. This allows evolution of the system, in a
manner that can be evaluated easily by anyone.
(4) to have a linguistic reflection capability so that people can ask
questions about the knowledge or structure of the system in English using
the defining vocabulary, and thereby learn specific facts about the system
itself without poring over hundreds of pages of documentation. The only
real "expertise" that the *basic* system needs is to understand its own
structure and an understanding of what people may need to know to use it
effectively, and be able to answer questions about itself.
(5) to have on-line an implementation of the latest version of that
language-understanding system (as there is for the MIT START system) so that
anyone can ask questions in English and get answers that allow one to
evaluate the current state of the language system, and determine whether it
is worth spending time to learn in full.
(6) to have a downloadable program that installs the ontology and associated
language-understandable components so that one can rapidly get up to speed,
asking questions and experimenting locally with additions to the ontology
and language components. (01)
Doug Lenat mentioned (if I interpreted him correctly) that he doesn't have
the funding to maintain an on-line demo of the language understanding
component. Without that, given the limited time available for most people
who may have an interest in the system, there is little motivation to spend
the time downloading and learning all the details needed to get the same
information locally. There is also very little opportunity for active
collaborative interaction and development. The history of Cyc as a
commercial program may have created an irreversible antipathy among
open-source enthusiasts that effectively blocks significant input from
anyone who is not a part of the official Cyc team. I believe that the
problem of language understanding is too large to be solved by any one team,
even one as large as Cyc, and needs to develop a collaborative mechanism for
incremental improvement of one or more common projects (each having a common
ontology and language interpreter) so that small improvements can be made by
isolated research groups. The current Cyc environment is not conducive to
that, even though they have made a lot public (which I use for my own
purposes) and have created a Cyc foundation to encourage the kind of
collaborative improvement I have in mind. It is not yet well enough
developed to allow experimental input without major effort. Perhaps it will
get to that point, and take off from there. Do you think that the activity
at http://www.cycfoundation.org/ suggests a vibrant collaborative community
for that project? (02)
The Cyc ontology might well form the basis for such a fully open system
(most of the COSMO is derived from Cyc). It needs to be fully free and open
- this may not be possible as long as Cyc is a commercial enterprise that
needs to make money to survive. If it changes to fully open, it may get
more traction, and I may not need to continue a separate project. I have no
idea what business model would permit Cyc to survive that way. (03)
It is my opinion based in unsupported intuitive evaluation of multiple
observations which I shall not reference, that very few will have the
motivation to spend the time needed to understand a system as complex as Cyc
unless they can see, with little or no effort, some demonstration that the
ontology enables applications that look more interesting than just a
database query; and that developing such applications is probably ten times
as difficult as building the ontology that powers the application. So we
don't have anything public that provides a lot of motivation to dig into any
of the existing systems, Cyc included. But that says **nothing at all**
about the virtue of the underlying ontology. What is says is that even if
ontologies are be essential for powerful semantic systems, they are only a
small component. That is, to me, one lesson of the Cyc experience. (04)
So to build the powerful ontology-enabled systems, either someone has to
come up with a lot of money for a coordinated project, **focused on the
basic language** (not military info or newspaper text), or a substantial
community has to adopt some foundation ontology as their common knowledge
representation, and incrementally build systems that grow collectively in
power by communicating complex information with each other via the common
standard of meaning. Cyc may play a role in this, but it has a history to
overcome and a lot more public relations work needed to overcome it. (05)
The basic components of SUMO are more open that Cyc, but SUMO has the same
problem of no public impressive application based on it. (06)
Meanwhile, I think it is useful to determine whether a foundation ontology
can be structured to serve as a Conceptual Defining Vocabulary, because that
kind of ontology may be sufficiently more open and easier to learn than Cyc
so that it might provide a lower barrier to entry for potential users that
don't have time to master the existing Cyc, or are put off by its still
partly proprietary nature, or its closed development method. (07)
Pat (08)
Patrick Cassidy
MICRA, Inc.
908-561-3416
cell: 908-565-4053
cassidy@xxxxxxxxx (09)
> -----Original Message-----
> From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-
> bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F. Sowa
> Sent: Sunday, March 30, 2008 12:39 AM
> To: [ontolog-forum]
> Subject: Re: [ontolog-forum] What is "understanding" - was: Building on
> common ground
>
> Pat and Pat,
>
> I agree with the comments by Pat H. The word 'understand' is
> sometimes used in a metaphorical sense; e.g., "This system
> understands Python." But when we're talking about AI systems,
> it's best to avoid loose metaphors and focus on exactly what
> kind of processing the system is doing.
>
> PC> And I do want to use an ontology that can support human-level
> > understanding. So that requires not only building the ontology,
> > but building the language processing system that will demonstrate
> > that it is useful for that purpose.
>
> As I've said before, Cyc is the biggest and most extensively
> tested and developed formal ontology on the planet. What would
> your proposal do that Cyc cannot do today? I don't know anyone
> who would claim that Cyc has reached anything similar to human-
> level understanding. What are you proposing that is different
> from Cyc? How would it accomplish what Cyc has not?
>
> PC> I haven't figured out how to do a few person-centuries of work
> > in my spare time, but language understanding is the ultimate goal
> > that *orients* the work.
>
> Cyc has done several person-centuries of work, and the ability to
> augment their knowledge base by reading the same kinds of books
> that people read was always one of their goals. They have had
> many top-notch linguists working with them. Yet they haven't
> achieved that. What are you proposing that they haven't done?
>
>
>
>
>
>
>
>
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-
> forum/
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/
> To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
> (010)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (011)
|