SUO: RE: Re: single vs. multiple ontology standard
Matthew,
At 04:16 PM 9/12/2001 +0200, West, Matthew R SITI-GREA-UK wrote:
>Dear Adam,
>
>See comments below.
>
>
>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.r.west@is.shell.com
>Internet: http://www.shell.com
>
>
> > -----Original Message-----
> > From: Adam Pease [mailto:apease@ks.teknowledge.com]
> > Sent: 10 September 2001 17:54
> > To: pat hayes
> > Cc: standard-upper-ontology@ieee.org
> > Subject: SUO: Re: single vs. multiple ontology standard
> >
> >
> >
> > Pat,
> > Here's the message that I sent to John that you were cc'ed on that
> > summarized previously stated arguments for a single ontology.
> > Just to
> > note, any standard has conformance requirements, and this need not be
> > framed as "forcing" anything.
> >
> > Adam
> >
> >
> > >Date: Thu, 06 Sep 2001 20:41:35 -0700
> > >To: "John F. Sowa" <sowa@bestweb.net>
> > >From: Adam Pease <apease@ks.teknowledge.com>
> > >Subject: Re: single vs. multiple ontology standard
> > >Cc: mcdavid@us.ibm.com, rekent@ontologos.org,
> > >James.Schoening@mail1.monmouth.army.mil, phayes@ai.uwf.edu
> > >
> > >John,
> > > There are several reasons, some of which are reasons why a single
> > > ontology might be considered better, others are evidence that it is
> > > possible (or at least not impossible).
> > >
> > >1. A standard with many choices is necessarily a less useful
> > >standard. Instead of building tools that support one data
> > model, multiple
> > >models have to be supported in order to be compliant.
>
>MW: This simply is not necessarily true. In fact it is ony true if the
>choices made happen to be the right ones for your application.
But isn't it a problem if one application chooses to be compliant with
module A and another chooses to be compliant with (alternative) module
B? Doesn't that mean that we have a weaker standard? In fact, doesn't it
mean that the standard is just the mappings and not the ontology modules
themselves?
>MW: You make (as meny implementors do) issues of the costs of
>implementation.
>On the other hand I would want to emphasise the ability to support business
>requirements. There is no good being cheap, if you can't do the job. It is
>not even the primary criterion.
I agree with all those statements but they could be used to support either
position.
> > >
> > >2. People have to make more choices with a multiple ontology
> > >standard.
>
>MW: I don't see being able to make choices as bad. Ask anyone choosing a
>car.
>The importatn things is to be able to make those choices that enable you to
>get the business job done.
Making choices isn't necessarily bad of course. The question is whether
making these choices is necessary. Different folks have different intuitions.
> > >In order to choose intelligently they should
> > understand all the
> > >alternatives and when to apply them (this is a weaker
> > argument than the
> > >one above because it's going to be a significant job just to
> > learn to
> > >apply one ontology right, but, just the same...)
>
>MW: I think you may overestimate who is going to be using these ontologies
>at the level of making choices. These choices will be related to business
>functions/purposes, and will be embedded in systems, rather than choices
>people make at run time. "ordinary users" should probably not even be aware
>that there is an ontology, just as they are probably not aware today of the
>data model for a system.
I agree that ordinary users won't be aware of this. I am also speaking
about system designers and data modelers. On 3d vs 4d for example, I doubt
very much that that choice would be based on business functions or
purposes. The choice among the very general theories that a SUO would
provide would not be related to the application level. I suspect those
choices would be made more on the basis of familiarity with a particular
theory.
> > >
> > >3. Despite advertising to the contrary, Cyc doesn't not
> > contain any truly
> > >alternative theories or contradictions. In fact, the entire upper
> > >ontology is in one context (baseKB). Like or dislike Cyc,
> > there's an
> > >existence proof that it's possible to build a single
> > coherent upper ontology.
>
>MW: I am quite sure that a single ontology can be created. It will just have
>
>limited value.
Ok, at least you agree it's possible which is one step closer to my
position than Pat and John.
> > >
> > >4. No one has yet shown that there are two truly
> > incompatible and equally
> > >valid theories that we need to include. I realize that Pat might
> > >reasonably disagree with this, but we simply haven't reached the
> > >conclusion of our discussion on 3d vs 4d. Each of us takes
> > a different
> > >view on who has the burden of proof naturally.
>
>MW: I dispair. But if you are right and we set off down a modular approach
>the only outcome will be that we have a more managable single ontology. So
>what's the problem in accepting a modular approach?
I've said that we should aim for a common ontology and if we find issues
that are truly incompatible that we should then create an alternative
module. But let's not judge that a priori. If we don't try to create a
single ontology, but rather just create a bag of alternate theories, we
won't have a standard.
This is also an issue that I'd be happy to compromise on. If people are
willing actually to create those alternate theories, and show how they are
both incompatible with some part of SUMO, and yet are still useful, I'd be
happy. What I object to are those folks who have rejected SUMO simply
because it is a single ontology.
If someone develops a theory of time for example that fits with SUMO's
structural ontology, but conflicts with the temporal ontology, and shows
how that theory is useful, that would be a great result. If further
scrutiny shows that the rest of SUMO can be supported in a meaningful way
that is just as valid as the existing temporal ontology, then the new
theory of time should be included as an alternative.
Note that this doesn't mean that someone should just introduce some
isolated conflicting axiom, or just introduce some position and justify it
on the basis of "this is an axiom accepted by community X", but rather show
how it is a really useful alternative.
Adam
> > >
> > >Adam
> >
> >
> >
> > At 11:49 AM 9/10/2001 -0500, pat hayes wrote:
> > >>John,
> > >> First, please let's acknowledge the validity of each others'
> > >> views. There have been serious arguments for what you term the
> > >> monolithic approach.
> > >
> > >Adam, could you cite those arguments? You have always
> > steadfastly CLAIMED
> > >that a monolithic appraoch was of utility, but the only
> > actual ARGUMENT I
> > >am aware of is the observation that CYC uses that approach,
> > which is taken
> > >as a kind of existence proof that it is possible. And that,
> > pathetic as it
> > >is, is the only actual argument I have heard you, or anyone, use.
> > >
> > >This is not to impugne your right to hold to your opinion, Adam, to
> > >emphasise the point. The issue is over "serious arguments",
> > not opinions.
> > >
> > >>If more people on the SUO list want a modular approach then
> > they should
> > >>be free to pursue that. That is what I'm proposing, that
> > each set of
> > >>people pursue the approach they're interested in instead of
> > trying to
> > >>force their views on each other.
> > >
> > >Since the basic issue seems to me to be that the
> > 'monolithic' approach
> > >amounts to the imposition of a single view, and the
> > 'modular' appraoch
> > >seeks to find a mechanism to accommodate several views
> > without trying to
> > >force one of them on the others, this proposal of yours has a rather
> > >hollow ring to it.
> > >
> > >Pat
> > >
> > >---------------------------------------------------------------------
> > >(650)859 6569 w
> > >(650)494 3973 h (until September)
> > >phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes
> >
> > Adam Pease
> > Teknowledge
> > (650) 424-0500 x571
> >
Adam Pease
Teknowledge
(650) 424-0500 x571