Re: PROV-ISSUE-568 (hadRole-domain): domain of prov:hadRole [Ontology]

Luc,

On Oct 23, 2012, at 5:44 AM, Luc Moreau <L.Moreau@ecs.soton.ac.uk> wrote:

> Hi Tim,
> 
> Thanks for the explanation.  It is now clear that the document reflects what
> was intended.  I have a comment/question and suggestion.
> 
> 1.  We came across this issue by writing a converter from rdf to other representations of PROV,
>   and making use of the "ontology compiler" elmo.
>   The resulting code allows us to write
> 
>    ex:e2 prov:hadQualifiedDerivation ex:d1
>    ex:d1 prov:hadRole ex:r


I'm not familiar with how Elmo works.
I assume that Elmo is letting you specify prov:hadRole on any subclass of Influence?
If so, then I would suggest that the tool is misbehaving.
"auto-filling" subclasses of a domain can be a convenient trick, but it does not follow the semantics of rdfs:domain.


>   The appendix indicates that the multiple RDFS domains and ranges [RDF-SCHEMA] 
>   for a property are interpreted as an intersection.

Yes.

> 
>   Are you saying that these statements are *not* consistent with an OWL2 DL/Full profile of prov-o, but are consistent
>   with an OWL2 RL profile of prov-o?

I'm not making a distinction about consistency in any OWL profile; prov-o should be consistent in all OWL profiles.
All RDFS domains (i.e., the Influence and the union) are asserted in all profiles, but the unions are not recognized or applied in OWL-RL (according to OWL-RL's definition).

> 
> 2. Given the importance of this appendix, I find it really buried.  The document is designed to allow browsing, but in 
>    section https://dvcs.w3.org/hg/prov/raw-file/default/ontology/Overview.html#hadRole
>    there is no indication that one should read this appendix to fully understand prov:hadRole.


Agreed.

>    Can we add a forwarding pointer to the appendix next to prov:Influence, in the domain of prov:hadRole?


Good suggestion.

I added the rdfs:comment into the ontology http://dvcs.w3.org/hg/prov/rev/37b3fc2c8796

which appears at http://aquarius.tw.rpi.edu/prov-wg/prov-o#hadRole

Do you have any suggestions on how to word the comment?

Regards,
Tim


> 
> Thanks,
> Luc 
>     
> 
> 
> On 10/22/2012 08:18 PM, Timothy Lebo wrote:
>> prov-wg,
>> 
>> On Oct 8, 2012, at 8:54 AM, Provenance Working Group Issue Tracker <sysbot+tracker@w3.org> wrote:
>> 
>>> PROV-ISSUE-568 (hadRole-domain): domain of prov:hadRole [Ontology]
>>> 
>>> http://www.w3.org/2011/prov/track/issues/568
>>> 
>>> Raised by: Luc Moreau
>>> On product: Ontology
>>> 
>>> 
>>> The definition of hadRole in prov-o
>>> http://www.w3.org/TR/prov-o/#hadRole
>>> 
>>> lists 
>>> prov:Association or prov:End or prov:Generation or prov:Invalidation or prov:Start or prov:Usage
>>> in its domain, which is what prov-dm states,
>>> but also
>>> prov:Influence
>>> which is not compatible with prov-dm.
>> 
>> 
>> It depends on what is meant by "compatible".
>> 
>> The appendix at http://www.w3.org/TR/prov-o/#prov-o-owl-profile
>> lists the "OWL-RL violation" of hadRole's domain,
>> 
>> prov:hadRole rdfs:domain [ (prov:Association prov:End prov:Generation prov:Invalidation prov:Start prov:Usage) ]
>> 
>> 
>>  and follows by providing a more general assertion that suits (and informs) OWL RL:
>> 
>> prov:hadRole	rdfs:domain	prov:Influence
>> The appendix also clarifies in narrative the meaning of rdfs:domain that can be mis-interpreted in other modeling paradigms (and "prov-dm"):
>> 
>> The more general domain should not be interpreted as saying, e.g., "prov:hadActivity can be used with any prov:Influence", but as "Anything using prov:hadActivity is (at least) a prov:Influence".
>> 
>> The appendix also states that  "some property domains or ranges have also been defined with the closest common superclass for the classes in the [OWL-RL-violating] union"
>> 
>> 
>> Tim
>> 
> 
> -- 
> Professor Luc Moreau
> Electronics and Computer Science   tel:   +44 23 8059 4487
> University of Southampton          fax:   +44 23 8059 2865
> Southampton SO17 1BJ               email: l.moreau@ecs.soton.ac.uk
> United Kingdom                     http://www.ecs.soton.ac.uk/~lavm
>