Re: CG: Re: Some references about ontology and analogy
On Thu, 06 Jan 2005 11:54:54 -0500, John F. Sowa <sowa@bestweb.net> wrote:
> I received an offline response to my note by someone who
> liked the references, but who disagreed with my comparison
> of the Semantic Web to Cyc:
>
> > ... bracketing of SemWeb approaches with Cyc just doesn't
> > make sense - they're very different. A completely different
> > set of factors will determine how well the former does...
Yup, that would be me ;-) [1]
> It's true that Cyc started long before the SemWeb, but its
> target customers and applications were *identical*. The
> only difference is that Cyc had a head start, developed
> the technology much further, and demonstrated failure (at
> least a failure to be commercially successful) much sooner.
Last point first - the SemWeb may never demonstrate failure, so
'sooner' is not a reasonable comparison.The differences are *vast*.
The Semantic Web is a direct extension of the existing Web. Check the
layer cake [2], the lower half has been implemented. I can't find any
usable statistics, but the number of users it has are: lots; the
number of documents: lots, number of asserted statements: already
approaching lots. Regarding age (not that it matters much), but the
SemWeb isn't exactly a spring chicken, many of the ideas are there in
TimBL's early Web docs.
There may be significant overlap between target customers and
applications, but look at the goals:
"Cycorp's vision is to create the world's first true artificial
intelligence, having both common sense and the ability to reason with
it."
"The Semantic Web provides a common framework that allows data to be
shared and reused across application, enterprise, and community
boundaries. It is a collaborative effort led by W3C with participation
from a large number of researchers and industrial partners. It is
based on the Resource Description Framework (RDF), which integrates a
variety of applications using XML for syntax and URIs for naming."
Nothing fancy. Just making the information more machine-friendly,
wiring the Web together better.
> And by the way, I had once had high hopes for Cyc. When I
> was at IBM, I had recommended that IBM join the group of Cyc
> sponsors. The reply I got was that the management at IBM
> research felt that they had better uses for a half million
> dollars per year. In retrospect, I can't blame them.
I'm sure plenty of value has come from the time and money that has
gone in to Cyc. Whether it could have been used better elsewhere is
anyone's guess.
> If anyone disagrees, please look at the slides, including
> the discussion of the ANSI/SPARC conceptual schema of 1978,
> which was intended to support interoperability among multiple
> applications by defining a common ontology (which was then
> called a "conceptual schema"). There is absolutely *nothing*
> being done today with ontologies and with languages such as
> RDF, OWL, and SWRL that has not been tried in much greater
> depth and sophistication over the past 20 or 30 years:
There wasn't a Web back then. See also TimBL quote below.
> http://www.jfsowa.com/talks/challenge.pdf
>
> Everything that is being proposed today for the semantic web
> is already implemented in Cyc.
Except the Web part.
Just look at the slides:
> an ontology of 600,000 concept types, over two million axioms,
> and six thousand microtheories (i.e., specialized subdomains).
> From the very beginning, the Cyc technology was made available
> to major universities and corporate research centers, including
> companies like Bellcore, Eastman Kodak, Microsoft, and the major
> gov't and military agencies around the Beltway. I used to say
> that *nobody* was able to develop any commercially successful
> applications. I have recently been corrected by someone who
> mentioned a couple of small applications, none of which were
> sufficient to keep Cyc from major "downsizing".
Oh yes, and the Semantic Web is being developed primarily in the open,
and the ontologies created in a distributed fashion, rather than being
one big lawbook.
> Looking at what is being done today, I see much more funding
> being poured into the SemWeb on technology whose only difference
> from Cyc is that (a) it's much less developed, (b) it uses a
> much clumsier notation, and (c) it has many more monkeys typing
> randomly at many more typewriters.
(a) hardly works, because many of the ideas were derived from Cyc (and
other KR efforts)
(b) Turtle is quite nice, as is the graphic notation. RDF/XML is
pretty ugly, but people don't really have to look at that much, and
there's a lot more RDF/XML around than KIF (several million weblogs
produce it, for a start)
(c) a great many more monkeys - anyone that adds data to the Web (or
associated systems, like this email archive), increasing the chances
of Hamlet significantly ;-)
> In my opinion, the proposed goals and applicatieons of Cyc include
> everything proposed for the SemWeb as a subset.
I'd say Cyc missed out on the key piece, the massively distributed,
end-user friendly infrastructure.
If anyone can
> show anything about the SemWeb that suggests a greater chance for
> success than Cyc, I would love to hear it.
The Web.
And for that matter,
> if anyone can suggest a way of using Cyc to help make the SemWeb
> successful, I would also love to hear that.
Build interfaces. Map models. Expose as much data as possible, ideally
using RDF over HTTP. Integrate inferencing systems. For a start...
I'd better quote TimBl's material from 1998 [4]:
[[
A semantic web is not an exact rerun of a previous failed experiment
Other concerns at this point are raised about the relationship to
Knowledge representation systems: has this not been tried before with
projects such as KIFand cyc? The answer is yes, it has, more or less,
and such systems have been developed a long way. They should feed the
semantic Web with design experience and the Semantic Web may provide a
source of data for reasoning engines developed in similar projects.
Many KR systems had a problem merging or interrelating two separate
knowledge bases, as the model was that any concept had one and only
one place in a tree of knowledge. They therefore did not scale, or
pass the test of independent invention. [see evolvability]. The RDF
world, by contrast is designed for this in mind, and the retrospective
documentation of relationships between originally independent
concepts.
]]
...and there's been a lot of progress since 1998.
Cheers,
Danny.
[1] http://dannyayers.com/archives/2005/01/06/ontologies-analogies-and-knowledge-soup/
[2] http://www.cs.umd.edu/users/hendler/Image1.gif
[3] http://en.wikipedia.org/wiki/World_Wide_Web
[4] http://www.w3.org/DesignIssues/RDFnot.html
--
http://dannyayers.com