Re: Action-164 suggestions (again) for unnamed individuals *in addition* to bnodes

On 2 Jul 2008, at 12:21, Alan Ruttenberg wrote:

> On Jul 2, 2008, at 7:12 AM, Bijan Parsia wrote:
>
>> On 2 Jul 2008, at 12:00, Alan Ruttenberg wrote:
>>
>>> Fulfilling action-164, Here is a pointer to the email when I  
>>> first suggested we consider supporting both bnodes and unnamed  
>>> (for the user) individuals.
>>>
>>> http://lists.w3.org/Archives/Public/public-owl-wg/2007Nov/0214.html
>>>
>>> Snippet, omitting unrelated, but not current suggestions.
>>>
>>> It seems to me that we could actually support both.
>>>
>>>> Individual(type(owl:Thing)) could be a skolem
>>>> SomeIndividual(type(owl:Thing)) could be an existential.
>>
>> This imposes a tree like restriction on the first.
>
> Good point!

This was how it was handled in OWL 1.

>  The syntax would need to "name" the unnamed individual.
>
> So better would be:
>
> UnnamedIndividual([handle] type(owl:Thing))
>
> where handle is name for the individual that has scoped within the  
> ontology only.

Aka bnode ids :)

>>> That is, we could offer two forms of anonymous individual  
>>> constructors. Neither would specify a name. One form of  
>>> constructor would allocate a unique (to exceedingly high  
>>> probability) name. The other would use a bnode, interpreted using  
>>> the usual existential semantics.
>>>
>>> Tree-shape restrictions would hold for the existentials, but not  
>>> for the other kinds of anonymous individuals.
>>
>> This would make entailment undecidable.
>
> Why? The individuals created by the "UnnamedIndividual()"  
> constructor are not really anonymous - they are named, just not by  
> the user. So we have tree-shaped restrictions for the existentials,  
> as in OWL 1, and arbitrary graphs of named individuals, as before.

If Bnodes have existential semantics and they appear in arbitrary  
graphs patterns, then we're back to where we started :)

>> It also doesn't address the legacy issue that people use bnodes as  
>> if they were skolem. We could consider evangelizing not using  
>> bnodes in RDF when skolem semantics were meant, but I don't see  
>> that working well anytime soon without a *big* push.
>
> True. However my action wasn't to address this. It is certainly an  
> important element in our discussion of what decision to make, though.

To tie it back to your action, I suspect we might get push back from  
having pseudo-anonymous individuals that weren't bnodes. I don't know  
how much, though.

Cheers,
Bijan.

Received on Wednesday, 2 July 2008 12:43:23 UTC