Dear John, (01)
> Although I have been arguing against a single fixed ontology, I'd
> like to mention that my favorite approach to ontology is indeed a
> 4D version -- namely, Whitehead's process-based ontology. However,
> I'm sure that ANW would be the first to admit that the version he
> presented in _Process and Reality_ has many loose ends that must be
> resolved before it could be considered an effective foundation for
> a computational system. I sketched some of the ideas in my KR book,
> but there's much more that needs to be done. (02)
MW: We have done rather more than sketch some ideas with ISO 15926.
It is now supported by some 20+ systems.
> However, I also believe that it is extremely important for any
> ontology to support a smooth mapping to and from natural languages.
> One of the main reasons is that the overwhelming amount of available
> info now and for many years to come is stated in NLs or forms that
> have been derived or abstracted from NLs. (03)
MW: Yes. one of the main things we have been doing in recent years
is working out how to map natural language terms to a 4D ontology/
data model. We made significant progress on this with the corporate
data model we developed for Shell's Downstream business.
> Even more important is that most people think in terms of their
> favorite NL, and they will inevitably interpret any info presented
> in any other form in the terms of that NL. If that info cannot be
> mapped smoothly into an NL statement, most people will almost
> certainly distort it (probably incorrectly) into a form that does
> map to their NL. (04)
MW: Yes. The problem is always finding a common language that can
be shared. I still think 4D is best for this.
> As an example, the linguist Roman Jakobson once said "I can speak
> Russian in 17 different languages." Since he spoke and wrote quite
> well in many of those other languages (especially English, German,
> and French), he showed that such mappings are practical.
> Since most NLs tend to treat space and time in something like
> a 3+1 coordinate system, any 4D system will have to support
> such mappings in both directions -- at least locally. (05)
MW: Indeed. However, I think this is worth the effort.
> Some comments on your note:
> MW> For any unambiguous description of some situation, it is possible
> > to define a mapping into a *good* integating ontology, effectively
> > creating a copy in the terms and paradigm of the
> integrating ontology.
> > This enables information from different sources and
> perhaps different
> > or no foundation ontology to be brought together, so that
> they can be
> > reasoned over or queried as a whole. A lot of Business Information
> > systems aim to do just this.
> I would agree that a 4D ontology is probably better as an integrating
> basis, but as I mentioned in earlier notes, most messages, including
> NL comments and explanations, are local. That means that a
> to or from an NL would not have to be globally consistent.
> It could be
> something like approximating a spherical earth by locally flat planes. (06)
MW: The key point about messages, is that if they all conform to a
common ontology, maping in and out at the system interface. This
introduces consistency and improves maintainability of the interfaces.
> MW> If you can't do this, you need to fix your upper ontology.
> I suspect that the upper levels of any globally consistent ontology
> would probably be highly abstract and not easily translatable to
> colloquial NL statements. (07)
MW: Well if you think that saying things can be classified, classes
can have subtypes, and individuals can have parts is abstract, then
yes, but these do not seem very abstract statements to me. (08)
> That is one reason why I would throw away
> the upper levels of most terminologies, including WordNet. (09)
MW: I wish you would stop saying this. The real problem is people
putting the wrong things in upper ontologies, not there being
something wrong with them per se. (010)
MW: As I once said to one of our senior managers about a failed
project: the fact that something has been done badly does not mean
it cannot be done well. (011)
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 (013)
Tel: +44 20 7934 4490 Mobile: +44 7796 336538
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (015)