Re: PROV-ISSUE-226 (dgarijo): domain of the qualifiedDelegation property? [Ontology]

Daniel,

Is this still a problem?

If so, could you add an example to http://www.w3.org/2011/prov/wiki/PROV_examples that exercise the relevant constructs in question?

Perhaps:

http://www.w3.org/2011/prov/wiki/Eg-23-leader-acted-on-behalf-of-country
->
http://dvcs.w3.org/hg/prov/file/tip/examples/eg-23-leader-acted-on-behalf-of-country/rdf/eg-23-leader-acted-on-behalf-of-country.ttl


I think relaxing the current domain of hadActivity would address your concerns, so that we can reference the Activity in which an Agent acted on behalf (from a prov:Responsibility).



I added responses within to help the discussions along.

On Jan 20, 2012, at 10:28 AM, Provenance Working Group Issue Tracker wrote:

> 
> PROV-ISSUE-226 (dgarijo): domain of the qualifiedDelegation property? [Ontology]
> 
> http://www.w3.org/2011/prov/track/issues/226
> 
> Raised by: Daniel Garijo
> On product: Ontology
> 
> actedOnBehalfOf is a n-ary relationship,

yes.

> since you can relate 2 agents and an activity. I have named the n-ary relationship "Delegation",

It is currently named "Responsibility" to suit the DM's naming.


> since (as Khalid suggested), it implies the notion of responsability between 2 agents.
> 
> All the other qualifiedInvolvements have as domain an activity.

This is not true. The "qualifiedInvolvements" (e.g. qualifiedUsage) may have a range of EntityInvolvement, ActivityInvolvement, or AgentInvolvement depending on the range of the corresponding unqualified property.

e.g.
prov:used range is Entity
prov:qualifiedUsage range is EntityInvolvement (which references the Entity using prov:entity).


> We could have the activity as domain for qualifiedDelegation too, but in this case the activity is optional, so it might not be the best approach.

To suit the qualified pattern, we must choose _one_ of the N resources in the relation to be the "primary" one. The rule here is to choose the one being referenced in the unqualified triple [being reified], and to reference it with {prov:entity, prov:agent, prov:activity} - all other resources in the N-ary are referenced with "hadX" properties.

This is discussed perhaps poorly at http://www.w3.org/2011/prov/wiki/ProvRDF#4._Naming_style_for_prov:entity_prov:activity_prov:agent_is_RESERVED


> 
> If we decide that the domain is an Agent,

But yes, we do choose Agent. Per the rule above.

> then we would need a "hadQualifiedActivity" property to link the optional activity which one of the agents is controlling on behalf of the other.


prov:hadActivity was created to reference an Activity for Derivation.
I propose to relax the domain on this property to handle the problem you point out.



> The second agent could be linked with the existent hadQualifiedEntity.

The second agent is already linked. See the example to see how :cameron_as_prime_minister is linked to <http://dbpedia.org/resource/United_Kingdom>.

> 
> Summary: if we choose the first option, it would be a wrong modeling IMO. If we choose the second option, we would need an extra property and now one of the qualified involvements would have Agents as domain. I've commited the delegation section in the html doc, but without any examples until we agree on this.
> Thoughts?

I couldn't parse our your two options clearly. Your second option summary sounds about right. I suggest we reuse hadActivity to cite the optional activity involved in the Agent-Agent relations.

-Tim



> 
> 
> 
>