Re: owl:TopObjectProperty in property chains?

On 26 Aug 2008, at 16:50, Jeff Thompson wrote:

>
> Uli Sattler wrote:
> >>  In
> >> other words, suppose you have the OWL 2 axiom:
> >>
> >> ObjectProperty: childRelatedToBrother
> >>  SubPropertyChain: hasParent o owl:TopObjectProperty o hasBrother
> >>
> >> would that be the same as this rule:
> >> hasParent(x, y) ^ hasBrother(w, z) -> childRelatedToBrother(x, z)
> >>
> >> In other words, the parent of x does not need to be the same as  
> the brother of z.
> >>
> >
> > Now this example looks very strange indeed: could you explain to  
> us what the idea behind it is? Cheers, Uli
>
> I'm trying to fill out the table of combinations of variables for  
> rules
> which can be converted to axioms without variables.  You already have:
>
> hasParent(x, y) ^ hasBrother(y, z) -> hasUncle(x, z)
>
> and
>
> hasPerformer(x, y) ^ loves(y, y) -> hasPrimadonna(x, y)
>
> How about this one:
>
> hasParent(x, y) ^ ownsCastle(y, z) -> hasRichParent(x, y)
>

hm, in this case, I would rather add the following axiom:

hasParent  some (owns some Castle)  subClassOf
HasRichParent

or rather

(owns some Castle)  subClassOf Rich

hasParent  some Rich subClassOf
HasRichParent



> Notice that the consequent has (x, y), not (x, z) so that z is  
> unbound.  I think this
> can done by turning ownsCastle(y, z) into a class description for y  
> like OwnsCastle(y) with
> a someValuesFrom restriction on ownsCastle
>
> Class: OwnsCastle  SubClassOf: ownsCastle some owl:Thing
>
> Then the rule becomes one which can be converted to OWL:
>
> hasParent(x, y) ^ OwnsCastle(y) -> hasRichParent(x, y)
>

> You see what I'm getting at.  In general, I'm interested in the way  
> that
> "Rewriting Rules into SROIQ Axioms" turns
> rules with variables into axioms without variables.

it's described in the papers mentioned earlier...but I think have a  
question in mind but you don't want to go through the algorithm's  
details?

>  Is there other work which turns
> all the forms of rules into axioms without variables, even if they  
> aren't tractable OWL axioms?

hmmm, no: all approaches only work for certain kinds of rules, and you  
have to be even careful to check, when you translate a *set* of rules,  
that the result doesn't violate the 'regularity restrictions' of OWL2  
subproperty axioms...

Cheers, Uli

>
>
> - Jeff
>
>
>
>
>

Received on Tuesday, 26 August 2008 17:42:30 UTC