RE: SUO: Re: Registry etc.
Mike,
A minimal registry would be descriptive but not normative, where by normative I mean that substantive issues like inheritance would be determined on the content side of things. Of course if you don't worry about substantive things like subsumption, inheritance, etc, then you haven't really "merged" the content in any meaningful way. That would be something like a minimal registry. Adam Pease just wrote in about a minimal registry and mentioned his acquiescence to this kind of notion.
But actually if the "registry" (damn word) is really a disguise for saying that we're going to take X and Y and create the new Z SUO, then this seems to require that there is some specific problem with X and Y that Z will solve. If the idea is that Z is a tool for creating content, we already have content in SUMO and OpenCyc, it won't do for everyone's interests, but neither does it have to "do" much either (I should have mentioned this in my prior email). Damn, when I'm developing content I want to link to Process or Object or whatever and forget about it! If I want to do philosophy I go to my department. For system design you just need assurance that there aren't inconsistencies "up there"--the real work is in mid-level terms that describe your domain. Of course when you actually go to implement anything you'll smash it down into an RDBMS and look the other way. But that is orthogonal. I gotta stop writing these emails too!!
Best regards (friend),
Erik
Mike Pool <mpool@iet.com> wrote:
At 08:31 AM 6/6/2003 -0700, Erik Larson wrote:
Mike,
If a "single, monolithic ontology" won't suffice for all content developers with their various interests and contexts, then neither will a registry that organizes one, two or ten ontologies. This is just the problem with specifing a set of usable concepts for many purposes, and I think it is largely independent of considerations of using or not using something called a "registry." Now if there isn't a magic bullet (and I believe you don't think that there is), then how best should we move toward an SUO? I assume you reject the monolith approach, and accept some form of "scalable, modular framework."
Yes, but I should say again that I think we may be making too big a deal of the need to be flexible with respect to new reasoning approaches, see below.
If that framework isn't to have a registry for incorporating and managing its content, what should it have? If it is to have some notion of a registry, what's wrong with the current attempts to outline one?
Well, as I said, I don't want the SUO to end up being nothing more than a registry of largely overlapping ontologies.
Is it that foundationally similar ontologies would have to co-exist? Is that too redundant? But then what is the option? One proprietary standard is backed and (by hypothesis it won't suffice for all content developers) and you have to have some way of adding content to it anyway, and managing the content.
Yes, this is a worry, It would be nice to adopt a single convention for labeling events, attributes, collections, specifying role players, articulating causal interactions, etc.. You would need a way to add content to the SUMO. But you don't need a way to add redundant content to it.
In the end I confess that the distinction that eludes me is the one implicitly presumed (by you? or no?): that there is some stable coherent ontology that will do just fine and then there is this hodge podge "registry."
Truth be told, I do think that is more or less possible. This SUO, in my mind will truly be upper level. I'm not envisioning it being Cyc 2. If you need to represent fish one way and I need to represent them another, then the SUO should provide some of the basic tools for each of these distinct representations but the SUO itself should not have the concept "fish" in it. So your need to come up with a different kind of representation from the one that the University of Guelph uses in their fish ontology, wouldn't obviously have perturbations that will manifest themselves in changes to the SUO. However, I also concur that there are fundament!
al representational issues that should be decided by application, e.g., 3D-4D, Davidsonian approach to event representation, etc. and that the SUO content should reflect and allow for this.
Finally, is your worry that the registry is a "waste of time" because all of the work goes into matching disparate terms from disparate sources instead of making new content? Well, then leave it minimal, people can spend their time how they see fit (like in making content for their own purposes). But if there is a need to put more work into it, great. There will be a framework and toolset for doing this. Or am I missing something? Some deep fundamental problem with the 'registry'?
Yes, this is also a worry. What do you mean, leave it minimal?
Erik
"West, Matthew R SITI-ITPSIE" <matthew.west@shell.com> wrote:
- Dear Mike,
- OK, your note below makes your concerns clear.
- A registry is a means to an end, not an end in itself.
- Does that help?
- Matthew West
- Principal Consultant
- Shell Information Technology International Limited
- Shell Centre, London SE1 7NA, United Kingdom
- Tel: +44 20 7934 4490 Other Tel: +44 7796 336538
- Email: matthew.west@shell.com
- Internet: http://www.shell.com
- > -----Original Message-----
- > From: Mike Pool [mailto:mpool@iet.com]
- > Sent: 06 June 2003 13:26
- > To: Cathy Legg; Eric Peterson
- > Cc: standard-upper-ontology@ieee.org
- > Subject: RE: SUO: Re: Registry etc.
- >
- >
- >
- > At 07:04 PM 6/5/2003 -0500, Cathy Legg wrote:
- >
- > >On Wed, 4 Jun 2003, Eric Peterson wrote:
- > >
- > >> > Mike,
- > >> >
- > >> ! > > Please spare me the false dilemma and histrionics.
- > >> >
- > >> > This is not a false dilemma, and it is "history", not
- > >> > histrionics. Please look at the SUO archives to see what
- > >> > was happening. All that discussion is on the record.
- > >> >
- > >> > Without a joint effort, we go back to the status of a few
- > >> > months ago, which was a moribund email list with ZERO
- > >> > collaboration among the three groups.
- > >>
- > >> [ELP] I would vote for a simple joint effort, but the
- > language of motion
- > >> #2 entirely avoids talk of a collaboration goal. And it adds the
- > >> controversial registry.
- > >
- > >As someone actively engaged in development on a very large ontology,
- > >Matthew and John's arguments for the registry seemed to me
- > to be simple
- &!
- gt; >software-development common-sense. QA for ontologies is *extre mely*
- > >challenging - possibly more so than for any other form of software
- > >engineering - so the more supportive tools and processes for
- > this kind
- > >of work the better.
- > >
- > >I've been trying to follow the messages complaining about the
- > >registry idea but for the life of me I can't make mental
- > contact with the
- > >problem. Something about 'content' being managed by the
- > registry, but in
- > >my understanding, any ontology is to some degree 'content
- > all the way
- > >up'. That's why you have one. So yes, content will be
- > managed. Is that so
- > >wrong?
- >
- > Cathy,
- >
- > The issue is that it's not clear what the function of the
- > registry would be, etc. I am, of course, all for QA,
- > motherhood, the war against terrorism and baseball on sunny
- > afternoons. The concern is not a concern with conte! nt being
- > managed by the registry, I'm not sure which messages you
- > read, it is an issue of what exactly the point of the
- > registry is with respect to the SUO objectives and how much
- > diverse content it will manage. If it is just an end in
- > itself, i.e., the SUO is a registry, it represents little
- > progress in the efforts to come up with a SUO over and above
- > efforts like those that can be seen at the Bateman URL to
- > which John has pointed us. If it is a tool for intricately
- > interweaving all kinds of contributions, each of which has
- > content that significant overlaps (but, of course, with
- > slightly different labeling and representational approaches),
- > with other contributions to the SUO, without trying to
- > significantly vet and choose from amongst different repre!
- > sentational conventions, then I see it as requiring very
- > high maintenance but being of very! little service.
- >
- > The ensuing discussion suggested tha t the proposed registry
- > is many things to many people, so I'm still not sure whether
- > the above concerns were justified. Or, one might say that
- > since the function of the registry with respect to the
- > ultimate SUO objectives just isn't clear to anyone, my
- > concerns about it are well-founded.
- >
- > John's recent restatements have made me feel a lot better
- > (but largely because there is no mention of a registry).
- >
- > best,
- >
- > Mike
- >
- > >> I tried to get a simple clause added to motion #2 that expressly
- > >> specified a goal of collaboration you said that it would
- > blow apart the
- > >> coalition.
- > >>
- > >> Adam, would your team leave if motion #2 stated that it's
- > chief goal was
- > >> a third merged ontology that seamlessly coalesced and
- > improved the two
- > >> content ontologies?> >>
- > >> John D., Cathy, Fritz, would Cyc leave the table?
- > >
- > >I wouldn't personally wish to, I can't speak for Cycorp.
- > >Cathy.
- > >
- > >-------------------------------------------------------------
- > -------------
- > >Cathy Legg, Phd Cycorp, Inc.
- > >Ontologist 3721 Executive Center Dr., ste 100
- > >www.cyc.com Austin, TX 78731-1615
- > >
- > > download OpenCyc at http://www.opencyc.org
- > >-------------------------------------------------------------
- > -------------
- >
- >
- >
Do you Yahoo!?
Free online calendar with sync to Outlook(TM).
Do you Yahoo!?
Free online calendar with sync to Outlook(TM).