Dear Pat, (01)
> Matthew,
> Just one point of clarification on the CDV hypothesis:
> [PC] > > I have said as part of this hypothesis that the required
> > > supplementation
> > > for any given domain will be small, once the basic CDV is on
> > > hand. But that
> > > is also part of the hypothesis, which needs to be proven
> > > (probabilistically
> > > - it can never be absolutely proven) by a serious attempt to
> > > build such a
> > > foundation ontology. And then by applying it to multiple domains.
> >
>
> [MW} > MW: As long as you are not insisting on there being no
> supplementation,
> > (which I had understood) then you are safe, and indeed,
> there are lots
> > of things that can be reused across domains - that is what upper and
> > mid-level ontologies are for. But now we are a long way from 6000
> > perhaps reducing to 4000, we are on a steadily climbing
> curve without
> > an asymtote.
> >
> The test that I have in mind is to represent domains using a basic
> foundation ontology, trying to represent the concepts of each
> new domain
> using what is in the existing foundation ontology, and adding
> new concepts
> to the foundation whenever it appears that new primitives are
> required to
> represent the domain. The domains are represented
> sequentially, and for
> each domain we learn how many new primitives must be added
> for each 1000 new
> concepts represented. The fraction of new primitives to
> newly represented
> concepts should decrease as each new domain is represented,
> but that is not
> enough to provide good evidence for the hypothesis. We will
> need to observe
> the rate at which the fraction of required new primitives
> decreases, and
> from that will (eventually) be able to estimate whether or
> not the total
> number of primitives will eventually reach an asymptote at
> infinity. The
> evidence will, of course be probabilistic since we can't be
> certain that the
> next domain will not require a much larger fraction of primitives. (02)
MW: Well my own experience is that you can get as much as 75%
reuse in related domains, and of course the more domains you cover
the more likely it is that the next one will be related. However,
this is still a long way from 100%. As to whether the asymtpot is
at infinity, one nearly needs to ask the question: is there
an infinite number of classes that can be constructed? Answer: yes. (03)
MW: So the only question is whether you decide to stop somewhere
and say "that is enough". But then you have to ask "Why would
everyone else agree on that point?" (04)
Regards (05)
Matthew West
Reference Data Architecture and Standards Manager
Shell International Petroleum Company Limited
Registered in England and Wales
Registered number: 621148
Registered office: Shell Centre, London SE1 7NA, United Kingdom (06)
Tel: +44 20 7934 4490 Mobile: +44 7796 336538
Email: matthew.west@xxxxxxxxx
http://www.shell.com
http://www.matthew-west.org.uk/ (07)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (08)
|