[DL] dl systems with concrete domain

[Carsten Lutz]

On Wed, 23 Apr 2003, Louis.Chevallier wrote:
>
> Sorry for the naive question, but it is buzzing me for so long time:
>
> Is there any real DL systems currently available that implement
> concrete domain together with reasonnably expressive logics and that
> allow "attribute chains".  

No. I believe that there is no DL reasoner at all that supports concrete
domains with attribute chains (except perhaps for some experimental systems
that have been abandoned long ago).

> I am thinking about seemingly simple need like the need of stating
> that any father is older than his son(s) A statement that would
> involve a hasAge real attribute.  

Yes, and attribute chains do not even suffice since I suppose that "son"
is not functional (and hence not an attribute). So you would need a
sequence "R g" with R role and g (concrete) attribute. 

> Racer does propose real attribute but it seems that some mysterious
> difficulties prevent me from entering such constraints
>
> Any clue?

Racer does not support (attribute or role) chains with concrete domains.
It is thus not suprising that you have problems with the above example.

You would need a logic such as the one described in

Carsten Lutz, Adding Numbers to the SHIQ Description Logic-First
Results. In Proceedings of the Eighth International Conference on
Principles of Knowledge Representation and Reasoning (KR2002). Morgan
Kaufman, 2002.

This logic allows to model your above example in a straightforward way.

But you want an implementation and the logic described in the above paper
has not (yet?) been implemented.  The main hindrance is that all known
decidability proofs for this logic are automata-based whereas modern DL
systems are tableau-based. I have some plans to try to find a tableau 
algorithm for this logic in the near future, and perhaps some of the system 
developers would like to take it up and extend their systems accordingly 
(Volker from RACER indicated interest in this). But that's dreams of the
future...

greetings,
		Carsten

--
*      Carsten Lutz, Institut f"ur Theoretische Informatik, TU Dresden       *
*     Office phone:++49 351 46339171   mailto:lutz at tcs.inf.tu-dresden.de     *