Re: On adding easy keys and top/bottom to the spec.

On Mittwoch, 30. April 2008, Achille Fokoue wrote:
> Hi,
>
> I was not able to attend last week's meeting where top/bottom properties
> and easy keys were discussed. According to Alan and the minutes of the
> meeting, there seems to be some consensus on adding them to the spec.
>
> On easy keys, we find them useful and indeed easy to implement using DL
> Safe rules. So I have no objections in adding them to the spec. In terms
> of impact to existing profiles, I assume they can be easily added to
> OWL-R, but should be excluded from DL Lite.   How about EL++?
>
> On top and bottom properties, I am not completely convinced by use cases
> given at [1]. Although use cases presented in Markus?s paper "All
> Elephants are bigger than all mice" are more compelling, they are use
> cases for concept product ? not universal property, which is just a
> special case of concept product.  I understand that they do not affect the
> worst case complexity of reasoning, but do we have any implementation
> experience about their practical impact?

It's true that concept (cross) products are a generalisation of the top 
property (since TOP x TOP = TopProperty), but, on the other hand, the top 
property can also be used to model concept products in a 
not-quite-so-cumbersome way (so one can claim use cases carry over):

Elephant = EXISTS elephant.Self
Mouse = EXISTS mouse.Self
elephant o topProperty o mouse  SUBPROPERTYOF  biggerThan

(I really should learn that Manchester syntax instead of making my own each 
time ...)

Not sure if the above works well with current reasoners (I suppose a reasoner 
could be much more efficient in dealing with Elephant x Mouse -> biggerThan 
directly). It also remains to be seen whether this will actually be a common 
use case in practical OWL 2 ontologies. It's not an overly obvious modelling 
pattern after all.

It should also be noted that the topProperty is often considered to be 
non-simple, so that it cannot be used with cardinality constraints directly. 
This should simplify the implementation. (I remark that this restriction is 
probably not really necessary; I guess that concept products can be ignored 
when considering restrictions like that of "simplicity" or "regularity" -- as 
in the case of rules -- ISSUE 22 -- this is the reason why they are "not 
quite" syntactic sugar only).

In any case, I think the explicit introduction of a name for the top property 
also serves ontology engineering (I remember tool builders asking for a name 
for the "top property" at some OWLED). 

Regards

Markus

>
> Thanks!
> Achille.
>
> [1] http://www.w3.org/2007/OWL/wiki/Universal_Property
> [2] http://www.korrekt.org/page/All_Elephants_are_Bigger_than_All_Mice



-- 
Markus Krötzsch
Institut AIFB, Universität Karlsruhe (TH), 76128 Karlsruhe
phone +49 (0)721 608 7362          fax +49 (0)721 608 5998
mak@aifb.uni-karlsruhe.de          www  http://korrekt.org