Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: SUO: RE: SUO E-mail Ballot -- Please Acknowledge receipt




John,

At 11:27 AM 7/21/2001 -0400, John F. Sowa wrote:
>Adam,
>
>This is a point I have repeatedly urged throughout the NCITS T2
>ontology workshops (1996 to 19997) and throughout the SUO efforts:
>the methodology is far more important than any particular collection
>of categories and axioms.  In fact, if your methodology is good enough,
>it should be able to create the axioms and categories automatically.
>That is the only cost-effective way to produce something the size
>of Cyc or larger.

Well, we're not creating axioms automatically and I don't see how that is 
reasonably possible with today's technology.  Methodology is good, but just 
because one doesn't have a formal one, doesn't imply that the result isn't 
good.

> > Ian has tried several times to explain, as best he can, what his
> > development process is.  How would you describe writing a novel?  Or
> > producing a UML design?  There is no cookbook procedure to developing a
> > large ontology.
>
>The answer is simple:  the most successful novel writers, UML designers,
>and programmers definitely follow a systematic methodology.  For novels,
>look at Isaac Asimov's list of over 500 books.  And for anything
>computer related, including ontologies, I would strongly recommend
>Fred Brooks' _Mythical Man Month_.

Much as I like Issac Asimov, I wouldn't use him as a paradigmatic case for 
a novelist!  As a physicist turned novelist, and one not of Literature, I 
wouldn't be surprised if he used an outline, graph paper or all sorts of 
sensible engineering tools which would be foreign to many other writers.

I also like Brooks' book and recommend it to anyone in my group who wants 
to be a senior level engineer.  Brooks has some excellent tips and 
guidelines but I wouldn't say he has much of a methodology.  But I think we 
digress.  The point is really

1.  Does SUMO need a codified methodology to be valuable?
2.  Given our description of the "in-practice" method for SUMO, is it good 
enough
3.  Do we have some special methodology that isn't being shared with the group?

I'd respond, no, yes and I wish we had something so valuable that it was 
worth keeping quiet (we don't, and my nature would be to share it even if 
we did).

> > We're doing our best to create (as well as explain)
> > something that to our knowledge has only been done once before on this
> > large a scale.
>
>I regard the Cyc approach as a proof that ontology development cannot
>and should not be attempted on a large scale without a solid, machine-
>aided methodology.

I guess we have differing opinions on Cyc.

> > It should not be a mystery that it is difficult to explain.
>
>There is an enormous literature on the subject, including the many
>years of accumulated experience in Cyc, WordNet, and other projects.

I'm not familiar with any paper on Cyc methodology.  I'm not as familiar 
with WordNet methodology as I guess I should be.  Could you recommend 
references?


> > The best explanation should be in reviewing the different
> > versions of the SUMO, and comparing them with the sources that have been
> > cited in the comments on each version.
>
>That is the work that should be documented by the people who did
>the job.  They know what they have done, and they should be able
>to put the comments together in a coherent HTML article that says
>exactly what they have done, why they did it, what the results were,
>and what more they plan to do.  And there should be concrete examples
>taken from the actual work to illustrate their points.

Ian and I did write a paper for my IJCAI workshop that makes a start at 
this.  We're working on more extensive papers as well.

> > A new version has been posted
> > publicly at least every two months, and often at an interval of a
> > week.  Ian has also cited on many occasions changes that he's made between
> > versions, as well as alerting the list to areas he plans to work on.  If
> > you don't like the product, that's one thing, but please don't impugn the
> > openness or intent of the process employed.
>
>I am not impugning the openness or the intent.  I am just asking for
>a better explanation than a log of changes from one version to the next.

I appreciate that John.  It seemed to me that Chris Partridge's comments 
attributed negative intent, but I know you haven't.  We'll keep doing our 
best to explain.

Adam


>John Sowa

Adam Pease
Teknowledge
(650) 424-0500 x571