[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

Re: [ba-ohs-talk] Re: Semantic web meta data

Eric Armstrong wrote:    (01)

> Jack Park wrote:
>>Of great value to OHS-think...
>>>>-----Original Message-----
>>>>From: p2p-hackers-admin@zgp.org
>>>>I have revised and expanded my decentralised meta-data
>>>>strategies document to include more thoughts on the issues
>>>>and some more systems (JXTASearch, SIONet, Reptile, Semplesh).
> This reminds me of an XTM question.
> Since a "topic" construct points to a URL to specify it's subject,
> what's to prevent multiple topics with differing names from pointing
> to different URLs, yet with the same meaning intended?    (02)

You pose this as a problem, yet it isn't necessarily one at all.
Or at least not a solveable one.    (03)

The idea that there is only one definition for any given thing
strikes me as more a horror than a benefit. What does anyone mean
when they say a word? How does anyone else interpret that word? One
of the reasons I joined the PORT list and have been digging into
the works of Charles Sanders Peirce (and of course others) is
because it's a fallacy to believe that such a thing as a "global"
ontology makes any sense whatsoever. Such an ontology could only
be coercive and incorrect. Dictionaries are not "correct"; they
only describe generalized usage, and they change constantly. Look
up the word "nigger" and tell me that it describes how one men
might greet another. Good luck if you don't know the context, or
know the men. And being white, I'm not even supposed to utter the
word in polite company. Does the dictionary state that? It says
"offens" but does it say *how* offensive? Then again, in certain
circles it's *not* offensive. Gad.    (04)

> The answer, of course, has to be standards -- an ontology standard,
> to be precise. But which? Cyc? Dublin core? other?    (05)

I'm not so sure. A notation can be standardized, but not the way
people think.    (06)

Cyc is a common sense ontology, that if you talk to Doug Lenat he'll
admit is basically "Doug's Ontology" (or perhaps "Cyc's ontology).
It's not global, universal, and I'm certain that anyone looking into
its structure will disagree with some part of it. That's natural.
Atheists will differ on definitions from Christians, eg., should
"Man" be classifed as an "Animal"? Are angels real?    (07)

Dublin Core is only a metadata expression, not an ontology, just
as is DAML+OIL, etc. ie., just ways of expressing ontologies. IEEE
has an ontology, as does Google, Yahoo, the US government (in
specific domains), etc. The US Federal Govt. has an XML working
group that is watching the W3C's progress and is aware of Dublin
Core. Unless I'm mistaken, Congress just passed some sort of bill
requiring that government documents be specified in an open format,
though I can't find reference to that now. If this is true, they'll
need an effective way of categorizing those documents.    (08)

> The goal I see for the semantic web, as Newcomb expostulates so
> eloquently in his chapter of the Topic Maps book, is for a user
> (me!) to get the information desired, while hiding the bushels of
> similar-but-not-desirable information.    (09)

Yes, this is what topic maps were intended to do. The principles
of topic map merging were meant to perform this magic, but one
must note that merging behaviours *can* be incorrect, as topic maps
(just like the rest of the known world) are imperfect. Humans (as
Steve Newcomb will tell you) are necessary in the loop.    (010)

> That tends to argue for natural-language queries. Otherwise I'm
> going to have a huge ontology to wade through to figure out how
> to express my question.    (011)

No, hopefully you'll simply be able to do this in natural language.
If the computer is unclear of your intent, it'll query you for more
information. I saw a demo of this in 1979, interestingly enough, it
was a Doug Lenat project there at SRI.    (012)

> Now then, when that query is processed, it will only be possible
> to use it sites that have somehow been tagged with the appropriate
> meta data. Of course, that will take a lot of work, so one needs to
> know it will pay off... (chicken & egg, since people will search with
> such a system *because* it will pay off.)    (013)

This is the premise of the W3C's Semantic Web effort. I was fortunate
enough to sit in on a session at ICCS last year where Terry Winograd
(to my mind) demolished the idea of such widespread use of metadata.
Only a very small portion of the web will be effectively so tagged.
But never fear, the web's search engines are ready to answer the
challenge. (See http://ftrain.com/google_takes_all.html)    (014)

> So one question is, who is focused on the www-o (world wide web
> ontology) problem?    (015)

There's a lot of different groups, all seemingly in competition,
unfortunately. The W3C has their WebOnto group, there's still I believe
the IEEE SUO (Standard Upper Ontology) list, there's some private
efforts like Cyc, and then there's that whole DAML+OIL thing (which
is both part of and independent of the W3C effort, being funded by
the US military, Stanford University, etc.).    (016)

> Another question is whether it might make even more sense to use
> on-the-fly network analysis to identify "groups", and let the zero in
> on set unions, intersections, & differences using "like this" and
> "not like this".    (017)

This is pretty commonly part of most ontology languages that I'm
aware of, certainly part of Cyc, DAML+OIL, OCML (here at KMi), etc.    (018)

As you might guess, I don't believe in global ontologies. They'd be
bad for the blood, bad for mental diversity, bad for thinking, just
bad to the bone. Now that doesn't mean I'm down on ontologies in
general. I'm all for grass roots, people-created, disagree'in,
confused and impossible-to-merge ontologies, which, like it or not,
are what we're going to get *anyway*. Ontologies are by nature local,
personal, time-, context- and domain-specific. Our vision of the world
is not static, nor formed from one person's (or one group's) mind.    (019)

Murray    (020)

Murray Altheim                  <http://kmi.open.ac.uk/people/murray/>
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK    (021)

      If it wants to be a global power and a player in the
      Atlantic alliance, Europe has to get back into the
      business of making war. -- Newsweek Magazine, June 3, 2002    (022)