Re: The Cannes Entailment [was: Coming to grips with the entailment put forth by Jeremy]

On 2002-04-29 10:41, "ext Brian McBride" <bwm@hplb.hpl.hp.com> wrote:

> At 08:45 29/04/2002 +0300, Patrick Stickler wrote:
>> On 2002-04-28 19:18, "ext Brian McBride" <bwm@hplb.hpl.hp.com> wrote:
>> 
>>> The know fix
>>> ============
>>> 
>>> In the know fix, we assume that the knowledge that
>>> 
>>>  <jenny> <age> "10" .
>>> 
>>> really means
>>> 
>>>  <jenny> <age> _:a .
>>>  _:a <xsd:decimal> "10" .
>>> 
>>> is defined elsewhere
>> 
>> This is almost but not quite correct. Unless you intended
>> to specify explicitly somewhere in the graph that
>> 
>>    age rdfd:datatype xsd:decimal .
> 
> Fair enough, I'm happy to add that as a precondition, though I don't think
> its really necessary in this case.

Well, it's necessary if the entailment is going to hold (IMMHO).


> [...]
>>> 
>>> The Rule Fix
>>> ============
>>> 
>>> We build a rule mechanism, as in the know fix, into RDF.  I think Jos has
>>> claimed that he has proved something similar, if not the same as this,
>>> works.  No one I have suggested this to has like it, usually for different
>>> reasons.
>>> 
>>>  <jenny> <ageA> "10" .
>>>  <ageA> <rdfd:datatype> <integer> .
>>>  <ageA> <rdfd:valueProp> <ageB> .
>>> 
>>> entails:
>>> 
>>>  <jenny> <ageB> _:a .
>>>  _:a <integer> "10".
>> 
>> There is an alternate way to fix this by rule without introducing
>> any variant properties. Namely the following closure rule:
>> 
>>    ?s ?p ?l .
>>    ?l rdf:type rdfs:Literal .
>>    ?p rdfd:datatype ?d .
>> 
>> entails
>> 
>>    ?s ?p ?x .
>>    ?x ?d ?l .
>> 
>> or
>> 
>>    ?s ?p _:a .
>>    _:a rdfd:lex ?l .
> 
> I was under the impression this one was subject to some criticism:
> 
>  a) Pat proposed it did not work (supported by Jeremy)
> 
>  http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Apr/0341.html

I'm still not convinced of that, based on a lack of clarity
about rdfs:Literal. See recent posts.

>  b) it has unfortunate interactions with things like
> daml:uniqueProperty.  What if ?p is unique?

This problem exists regardless of the closure rule. Even if
the closure rule doesn't entail it automatically, you can
still have explicitly asserted in the graph both

   Jenny ex:age "10" .

and

   Jenny ex:age _:x .
   _:x xsd:integer "10" .

which will result in exactly the same problem with either
daml:uniqueProperty or owl:UniqueProperty.

The problem is not with the closure rule. The problem is
that literals cannot denote a datatype value, and hence,
one can *never* use both the inline idiom and a bnode
based idiom in the same graph with a daml:uniqueProperty.

>> Why xsi:type?
> 
> Because that is what xml schema uses.

But then we have to adopt a good portion of XML Schema semantics
to define how it works.

>>  And why for every occurrence?
> 
> Every occurrence?

Well, I presume you'd have to put that attribute on every
property element to indicate the datatype of its object, no?

> 
>> The point of the inline idiom is that the datatyping is implicit
>> (global, not local). The above is just as local, and more verbose,
>> than the datatype property idiom:
>> 
>>      <rdf:Description rdf:about="#jenny">
>>         <age xsd:decimal="10"/>
>>      </rdf:Description>
> 
> True.  But the xsi mechanism may allow us to use DTD's and schemas to
> override the interpretation of existing RDF.

I don't see how the xsi mechanism works any differently from the
datatype property idiom, and it does so with a more verbose serialization
and requires additional definition/specification. The only way it
differs from the datatype property idiom is that it doesn't introduce
a blank node, which I guess would give us a symmetrical set of four
idioms:

              Blank Node       No Blank Node

Global         rdfd:lex          inline

Local          datatype         xsi:type


But do we want another idiom, and does not introducing a new
idiom also touch the stake in the ground?

Patrick 

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Monday, 29 April 2002 04:28:21 UTC