Re: [OWL Compatibility] Re: RIF & OWL compatibility

On 4 Jan 2006, at 08:59, Jos de Bruijn wrote:
>> What is a "superset approach"? Is it really true that "There are
>> several nonmonotonic logics which are candidates for such a
>> formalism, such as circumscription, autoepistemic logic and default
>> logic"? Can you encode DL's in such logics? Well, yes - if you really
>> mean a superset of FOL with non-mon operators; but does it really
>> make sense to say that? In the "black box approach", as you mean it,
>
> Yes, a superset of FOL was meant here. This approach probably doesn't
> make sense in a practical setting, but it is a possible approach.

I am not really sure that such a superset of FOL (capturing both the  
LP aspects and the DL aspects) exists and/or has been studied.  
Although we can not exclude that such a beast exists, it has unknown  
properties and therefore I would not even mention this possibility.

>> it is *not* true that "OWL DL and the LP rules are viewed as black
>> boxes which exchange only ground facts". Please, first study Rosati's
>> and Eiter's works before stating these things. Moreover, the
>> statement: "Second, the existential consequences cannot be reused,
>> since it is not possible to deal with existential information in LP
>> rules", is false.
>
> I actually defined the black box approach there, and I defined that  
> the
> OWL DL ontology and the LP rule base are viewed as black boxes which
> exchange ground fact. Perhaps this approach needs to be refined,  
> but I'm
> not sure.
> I'm not yet aware of Rosati's approach, but his paper is on my stack.
> As far as I understood Eiter et al's approach, it works exactly as
> described here. One can include queries to a DL ontology in the  
> body of
> LP rules; it is also possible to specify which facts needs to be added
> to the DL ontology before executing the query.

Things are more complex than this, and several different incompatible  
options are available. First of all, you are neglecting all the FOL- 
based approaches (e.g., SWRL, Carin, etc). If we stay in what you  
call LP semantics for rules, then the interaction between rules and  
the ontology would be better characterised by looking at how the  
models of the rules and the models of the ontology do interact,  
rather that by having ground facts exchanges.

> Wrt. the existential information: the semantics of LP rules is  
> based on
> Herbrand models, which means (among other things) that every  
> individual
> in the universe is represented with a name in the language. This makes
> it impossible to truly capture existential information.

I disagree. For example, In Rosati's approach non-distinguished  
varibales (i.e., existentials) are not restricted to the herbrand  
universe. BTW all the use cases in Managing Incomplete Information  
<http://www.w3.org/2005/rules/wg/wiki/ 
Managing_incomplete_information> would be correctly handled by  
Rosati's approach.

cheers
--e.

Received on Sunday, 8 January 2006 20:39:35 UTC