W3C home > Mailing lists > Public > www-webont-wg@w3.org > May 2003

Re: owl:All(something) ??

From: Ian Horrocks <horrocks@cs.man.ac.uk>
Date: Wed, 14 May 2003 23:54:43 +0100
Message-ID: <16066.51507.126969.415741@merlin.horrocks.net>
To: Jim Hendler <hendler@cs.umd.edu>
Cc: webont <www-webont-wg@w3.org>

On May 14, Jim Hendler writes:
> 
> At 10:14 AM +0100 5/12/03, Ian Horrocks wrote:
> 
> >As Jeremy points out, the various "same" relations are transitive, so
> >there is no n^2 blowup in the number of property relationships
> >required to express them, and the result is more compact than a (fully
> >expanded) list would be.
> >
> >I also agree with Jeremy's account of the WG's thinking
> >w.r.t. disjoint classes - that it is much less frequently required,
> >and that the numbers (of classes) involved would typically be small
> >(compared to unique individuals). It was therefore decided that
> >additional syntax was not justified.
> >
> >W.r.t. disjoint properties, I'm not sure we even considered it. It is
> >rarely asked for, and unlike the class/individual cases it is
> >relatively difficult to express in the existing language given
> >that we don't have property negation.
> >
> >Ian
> >
> >
> 
> 
> I had pretty much decided the above was correct, but yesterday I was 
> talking to scientists about Semantic Web, and the following use case 
> came up -- they want to be able to take biological taxonoma and 
> represent in OWL, but they care very much what is and isn't disjoint 
> (class wise).  In particular, they want to be able to have it be the 
> case that at each level of the taxonomy the subclasses are disjoint 
> -- i.e. they want to be able to say
> 
> AllDisjoint (Animal, plant, ...) 1s of classes
> 
> AllDisjoint (Mammalian, reptilian, amhibian ...)   (10s of classes)
> 
> and then
> AllDisjoint (canine, feline, ovine, cervine, ceatacean ...) (100s of classes)
> 
> but not to assert that within canines there is disjointness between 
> wolfes, dogs (of various types) etc, since these can interbreed etc. 
> They said the bottom level would be in the 100s of classes, so could 
> need 10000+ separate disjoint statements!
> 
> (Note that they used the animal example so I could understand, but 
> they were actually talking about more subtle distinctions in 
> proteomics and the like).
> 
> In short they had some compelling use cases in the same general area 
> as Alan Rector's (scientific use of OWL in medicine/biology) and are 
> purely at a class (not property) level, so maybe we need the syntax - 
> esp. as it would be nearly the same syntax as for the alldifferent 
> individuals.

I should mention that Alan is also pretty keen on asserting
disjointness for sets of classes (he claims that, in practice, sets of
sibling classes are very often disjoint).

Of course we already have the necessary abstract syntax, so the
problem is only with the mapping to triples.  Is this really
significant given that RDF syntax is already incredibly verbose?

Another random thought - if you are really worried about verbosity you
could express the disjointness using a functional datatype property
with different values, one for each disjoint class. Of course this
wouldn't be very pretty from a mapping perspective :-)

Ian


> 
> Note, in OWL full I would do this by simplying using the allDifferent 
> on the classes - maybe that is the solution to suggest to these folks 
> for now, although could mean a lot of scientistific applications 
> would not be using DL...
> 
>   -JH
> 
Received on Wednesday, 14 May 2003 18:54:57 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:58:00 GMT