RE: A NEW FUNDAMENTALLY DIFFERENT FORMAL MOTION: was RE: SUO: Re: SUO Ballot with 2 Questions

["West, Matthew R SITI-ITPSIE" <matthew.west@shell.com>]

Dear Eric,

There is one point below where we are seriously missing each other.

You seem to be under the impression that if we have an SUO which includes
different foundational approaches and a mapping between them, then any
external ontology would have to map to EACH of these.

On the contrary, an external ontology would only have to map to the one
which was MOST SIMILAR to itself. The internal mappings would handle the
links to the other foundational ontologies. This would simplify the
mapping between ontologies and reduce the chances of error.

On your last point the thing I notice in general is that a common reaction
when faced with a difficult problem is to look for something easier, and
pretend that will solve the real problem. It doesn't, it just puts off
the time when you realise you have to address the real and difficult
problem.


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: Eric Peterson [mailto:epeterson@CCAAVA.com]
> Sent: 13 June 2003 20:15
> To: West, Matthew R SITI-ITPSIE; Jean-Luc Delatre
> Cc: John F. Sowa; Mike Pool; apease@ks.teknowledge.com; clegg@cyc.com;
> John DeOliveira; Patrick Cassidy; standard-upper-ontology@ieee.org;
> semanticweb@yahoogroups.com; Downes, Stephen; cg@cs.uah.edu
> Subject: RE: A NEW FUNDAMENTALLY DIFFERENT FORMAL MOTION: was RE: SUO:
> Re: SUO Ballot with 2 Questions
> 
> 
> Thanks Matthew;
> 
> Please see below:
> 
> <snip>
> 
> > > > > It wouldn't make sense to interoperate in this way 
> with multiple
> > > > > competing ontologies.  You would need n^2 mapping 
> axioms to link
> n
> > > > > competing axioms for a given definition.  Or you 
> could map all n
> > > > > competing axioms into one interchange axiom, but them 
> you end up
> > > > > creating a single interchange ontology to link your competing
> > > > > ontologies.  One way or another, scalable interoperability is
> best
> > > > > supported by a single upper ontology.
> > > > >
> > > > > It's really that simple.
> > > >
> > > > MW: Well not really. All the systems out there also use 
> different
> > > > flavours of (usually implicit) ontology. So you have 
> quite complex
> > > > mappings to manage, many of which will be essentially duplicates
> > > > of each other.
> > >
> > > [ELP] Mappings, I claim, are irrelevant.
> > 
> > MW: I don't see how.
> 
> [ELP] I must not see your original point.  I agree with your two
> statements, but I don't see how they bear on the n^2 vs. interchange
> ontology hub question.
> 
> > 
> > > But I rather map into one
> > > ontology than 12.
> > 
> > MW: But you would only have to map into one - other views would
> already
> > be mapped.
> 
> [ELP] Would you really want a wad of O(n^2) mappings rather than an
> interchange ontology with O(n) mappings?
> 
> <snip>
> 
> > > I believe in altruism, but even if I thought that it was remotely
> our
> > > responsibility to map the world into our ontology, it 
> would view it
> as
> > > being one of the many things we don't have time to even 
> contemplate
> > > doing as a standards group.  We can't be all things to all people.
> At
> > > VerticalNet, one of our ontology support groups wrote software to
> map
> > > relational DB's into our ontology.  If a commercial endeavour was
> that
> > > eager to do a so much serious mapping work, why should we as
> > > a standards
> > > group be so eager to take it on gratis.  Folks will pay us well to
> do
> > > it.
> > 
> > MW: Our role is to provide the environment and foundation 
> (read upper
> > ontology) so that people can do it for themselves (or pay 
> others to).
> 
> [ELP] Then we agree.  
> 
> > >
> > > > On the other hand, if you cater for the major flavours
> > > > of ontology and map between them once, you greatly simplify the
> > > > mapping process for those wishing to integrate, and do 
> the hardest
> > > > parts for them.
> > >
> > > [ELP]  How is n^2 mappings per axiom simple?
> > 
> > MW: It doesn't have to be that bad, and when n is small 
> (2-4) does not
> > present a major problem. 
> 
> "Represent once" is one of my most fervently recited mantras.  You are
> advocating duplicative representation of mappings. 
> 
> > It is even likely that it will improve the
> > quality. Usually, ISO standards have to be written in at least 2
> > languages. The reason for this is that in doing the translation,
> > linguistic perspectives that represent ambiguity are uncovered.
> 
> [ELP] I claim that the hub model would afford ample opportunity to
> improve the upper model in the way that you describe.  
> 
> And if you want a second language, the hub is still connected 
> to each of
> the twelve other ontologies.  You can take your pick.  The 
> hub, however,
> avoids the other exponential hoard of connections between 
> ontologies as
> we scale up.  I, strongly feel that not addressing scalability in this
> way from the beginning would be a big mistake.
> 
> 
> > 
> > > And this group producing
> > > mappings as required product is completely outside of the charter:
> > 
> > MW: Where does it say "thou shalt not do mappings?" A mapping is
> really
> > just another word for implication. Are you saying our 
> ontology cannot
> > have implications in it?
> 
> [ELP] Are we discussing ontologies or upper ontologies?
> 
> I don't mind having a mapping entity/class in a UO, but 
> that's where it
> ends for me.
> 
> My threads of SUO hope are both getting pretty threadbare;^)  
> Every time
> we talk about another SUO deliverable, I vacillate between a sense of
> futility and of denial.  Do you ever worry about this group taking on
> way too much?
> 
> 
> <big snip>
> 
> [ELP] 
> Best,
> 
> -Eric
>