RE: Problems with dark triples approach

> Well, I have not yet seen a worked-out version of the solipsistic stuff.
> In particular, how are conditionally-existing classes handled in this
> approach.
> 
> Consider the following query:
> 	John rdf:type _:1 .
> 	_:1 daml:onProperty foo .
> 	_:1 daml:toClass daml:Restriction .
> 	John rdf:type _:2 .
> 	_:2 daml:onProperty foo .
> 	_:2 daml:hasClass _:3 .
> 	_:3 daml:OneOf _:4 .
> 	_:4 daml:first _:R1 .
> 	_:4 daml:rest _:5 .
> 	_:5 daml:first _:R2 .
> 	_:5 daml:rest daml:nil .
> 
> Which restrictions are to be added to the premises for this sort of query?
> Of course, it doesn't matter here, because neither _:R1 nor _:R2 are
> interesting, but a larger example could be constructed where it would
> matter.

Yes that's a nice example.

If I have understood, you have set it up so that either 
_:R1 or _:R2 is a daml:Restriction. Extending the example
and assuming the triples are not dark and the implicit
consequences carry the semantics then it is difficult
(read "not realistic with reasonable termination") to 
implement this. 

To some extent this
difficulty is the case whenever we set up a disjunction,
and users need to have some sort of expectation that not
everything that they might say is necessarily going to
be soluble within reasonable time.

I can see that examples where the disjunction is within
the ontology machinery are much more likely to lead to
excessive search space.

Jeremy

Received on Tuesday, 23 April 2002 10:21:43 UTC