RE: PROV-ISSUE-321 (dgarijo): Instances of involvements can be expressed without a subclass. [Ontology]

Sorry, this was kind of a train-of-thought email that went a little off
topic. 

:-)

--Stephan

-----Original Message-----
From: Stephan Zednik [mailto:zednis@rpi.edu] 
Sent: Wednesday, March 14, 2012 8:07 PM
To: 'Provenance Working Group'
Subject: Re: PROV-ISSUE-321 (dgarijo): Instances of involvements can be
expressed without a subclass. [Ontology]


On Mar 14, 2012, at 7:34 PM, Stephan Zednik wrote:

> 
> 
> -----Original Message-----
> From: Provenance Working Group Issue Tracker
[mailto:sysbot+tracker@w3.org] 
> Sent: Wednesday, March 14, 2012 6:37 PM
> To: public-prov-wg@w3.org
> Subject: PROV-ISSUE-321 (dgarijo): Instances of involvements can be
expressed without a subclass. [Ontology]
> 
> PROV-ISSUE-321 (dgarijo): Instances of involvements can be expressed
without a subclass. [Ontology]
> 
> http://www.w3.org/2011/prov/track/issues/321
> 
> Raised by: Luc Moreau
> On product: Ontology
> 
> The ontology allows for instances of involvements to be expressed, without
specifying its subclass (Usage, Generation, etc). This is not aligned with
the data model.
> 
> The only way to prevent this would be to define Involvement as a union
class, which is not allowed in OWL2-RL.  This would also prevent users from
extending Involvement directly, they would have to extend one of the
subclasses.  I am not sure if we want Involvement itself to be an extension
point in the ontology.
> 
> Also, I would say that this is not aligned with the PROV-N notation in DM.
The DM only defines PROV-N constructs for classes that would be leaves in
the concept hierarchy; this effectively means the PROV-N construct hierarchy
is flat - which is not something we are looking to replicate in the
ontology.

Well, looking at the latest DM I wouldn't say the DM has ~no~ concept
hierarchy, but I must admit I am not sure if the hierarchy has the same
implications it would in OWL/RDF.

1) Element appears to be an abstract concept since there is no DM construct
for it, so you can't assert it directly.
 
The only way to support this in the ontology would be to not have
prov:Element in the ontology, or to make prov:Element a union class (and
therefore break our OWL2-RL requirement).  Using a union class would make
more sense if we want to use prov:Element as the domain of
prov:hasAnnotation.

With either option, users would not be able to extend prov:Element since it
wouldn't exist or would be a union class.  I am not sure we would want this
anyway.

2) Agent is a specialization of Entity, but ramifications of this are not
detailed 
- can an Activity use an Agent?
- can an Activity generate an Agent?
- can an Agent be derivedFrom another Agent? of an Entity?
- can an Agent be an alternateOf another Agent?  of an Entity?
- can an Agent be a specializationOf another Agent? of an Entity?

These are all implications of Agents also being Entities in OWL/RDF, but I
am not sure if they are intended implications in the DM.

3) Relations are grouped in the document, but there is no clear hierarchy in
DM constructs.  Relation, Activit-Entity Relation, Activity-Agent Relation,
Entity-Entity Relation and Entity-Entity Relation are all abstract concepts.

I think this means the relation hierarchy in DM is flat.  

I believe a hierarchy was added in the ontology to both logically group in
the ontology as they are in the DM document structure and to make property
domains and ranges easier to specify.

--Stephan

> 
> --Stephan
> 
> 
> 
> 
> 
> 
>

Received on Thursday, 15 March 2012 06:59:33 UTC