W3C home > Mailing lists > Public > www-rdf-logic@w3.org > May 2004

Re: Concrete and abstract domains disjointness

From: Ian Horrocks <horrocks@cs.man.ac.uk>
Date: Tue, 11 May 2004 09:42:08 +0100
Message-ID: <16544.37344.91801.247454@excalibur.cs.man.ac.uk>
To: Jeremy Carroll <jjc@hplb.hpl.hp.com>
Cc: "Dimitrios A. Koutsomitropoulos" <kotsomit@hpclab.ceid.upatras.gr>, www-rdf-logic@w3.org


I imagine that you don't really want, e.g., the individual "John" and
the data value that is John's social security number to be the same
object; what you want is to be able to use data values as DB-style
keys. As Jeremy points out, this is possible in OWL-Full by using
inverse-functional datatype roles. In [1] you will find an analysis of
the consequences of adding a more comprehensive key mechanism to
description logics (there is also a technical report at [2] if you
want all the gory details). As you will see, there are decidability
and complexity issues, but adding keys to a DL is theoretically


[1] http://www.cs.man.ac.uk/~horrocks/Publications/download/2003/LAHS03a.pdf
[2] http://lat.inf.tu-dresden.de/~clu/papers/archive/ltcs02-04.ps.gz

On May 10, Jeremy Carroll writes:
> Dimitrios A. Koutsomitropoulos wrote:
> >  
> > 
> > However, because of the consequent seperation of object and datatype
> > properties, one still looses the (simple) ability to connect uniquely
> > individuals and datatypes. For example,given an ID number, one can never
> > retrieve the person to whom it corresponds to (using OWL DL of course, but I
> > doubt if this is possible even in Full). Whould this be so hard for a DL
> > reasoner to implement?
> > 
> It's possible in OWL Full using an InverseFunctional Datatype Property.
> I share your skepticism that this is not too hard for DL ... Ian 
> Horrocks has argued on a number of occassions that it is, but stubbornly 
> I remain unconvinced that this feature is truely harder than others that 
> are already in OWL DL. But I can't point to working DL style reasoners ...
> Jeremy
Received on Tuesday, 11 May 2004 02:43:45 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 2 March 2016 11:10:42 UTC