Re: PROV use for a Contributor Attribution Model

Hi Matt,

CC’ing in one of the DCAT (rev) editors, Simon.

> Curious about PROV's stance on such 'extensions'
I guess unless it breaks any PROV rules, the PROV editors can’t complain (can any of you??). The property dcat:hadRole does have this domain:

      a owl:Class ;
      owl:unionOf (
          prov:Attribution
          dcat:Relationship
        ) ;

But it’s got no range given and no axiom relating it to PROV property paths, even though these are sort of stated in the DCAt guidance document.

Even dcat:qualifiedRelation which says “Introduced into DCAT to complement the other PROV qualified relations.” doesn’t contains any PROV subclassing, relations via axiom etc. DCAT’s pretty loose…

> What are implications for interoperability with PROV-based data models?
When I implement DCAT (rev), I’ll be doing so via profiles (defined using the Profiles Vocabulary [1]) that will have strict PROV relations. Using the Profiles Vocabulary, I’ll link CONSTRUCT statements or SHACL files to DCAT to build out PROV things. This isn’t “perhaps”: I have a DCAT-based catalogue I’m working on now that I’ll do this for [2]

Cheers,

Nick


[1] https://w3c.github.io/dxwg/prof/
[2] https://horizon.gsq.digital/






From: Matthew Brush <brushm@ohsu.edu>
Date: Tuesday, 1 October 2019 at 4:55 am
To: Nicholas Car <nicholas.car@surroundaustralia.com>, "public-prov-comments@w3.org" <public-prov-comments@w3.org>
Cc: Sarah Ramdeen <sramdeen@ldeo.columbia.edu>, "thessena@oregonstate.edu" <thessena@oregonstate.edu>, "ddubin@illinois.edu" <ddubin@illinois.edu>
Subject: RE: PROV use for a Contributor Attribution Model

Hello, and thanks for the response Nicholas.  Yes, I have reviewed the RDA TDWG attribution model (Anne, who contributed to this work, is a colleague of mine). My understanding is that it essentially proposes the PROV structure - which requires 4 objects to represent an attribution scenario (the Entity, an Activity, an Agent, and a qualifiedAssociation between the Activity and Agent from which the Agents role can hang).  Given this, it doesn’t meet the need our project has for an even simpler 3 object structure.  That said, I have had discussions with Anne and hope to work toward alignment of our efforts. If you or other colleagues would like to participate in these discussions all are welcome.

As for the W3C Dataset Exchange WG work on an update to DCAT, thanks for the pointer here.  Looking at the link you shared, it seems this schema gets around the PROV constraint concerning the prov:hadRole property by simply defining their own dcat:hadRole property and using it as they like (i.e. hanging it from a Qualified Attribution directly). Curious about PROV's stance on such 'extensions' - did they endorse the specific solution DCAT ended up implementing? What are implications for interoperability with PROV-based data models?

In essence the CAM model is not much different in its approach - if we consider the cam:Contribution class to be a prov:Attribution for which we have defined our own version of prov:hadRole to use here, as well as our own versions of other PROV properties not formally allowed on prov:Attributions but desired for cam:Contributions (e.g. CAM versions of prov:hadPlan, prov:atLocation, prov:startedAtTIme, prov:endedAtTIme).  In addition CAM defines extensions for properties to hang things like organizational context funding source from the Contribution object.  Curious how PROV would view such an 'extension' of their core model, and if/how we might frame this as a proper endorsed PROV profile/extension?

Best, Matt

From: Nicholas Car <nicholas.car@surroundaustralia.com>
Sent: Monday, September 30, 2019 4:51 AM
To: Matthew Brush <brushm@ohsu.edu>; public-prov-comments@w3.org
Cc: Sarah Ramdeen <sramdeen@ldeo.columbia.edu>; thessena@oregonstate.edu; ddubin@illinois.edu
Subject: Re: PROV use for a Contributor Attribution Model

Hi Matthew,

Are you aware of the recent work by the Research Data Alliance / Taxonomic Databases Working Group for “Metadata Standards For Attribution Of Physical And Digital Collections Stewardship”? By the sounds of it there’s a bit of crossover here. I’ve CC’d in some of the relevant RDA / TDWG people.

Also, the issue you note about roles hanging on to objects was noted both in the PROV WG, recently here on this list (by me!) and is addressed somewhat in the current W3C Dataset Exchange WG work on an update to DCAT, see https://w3c.github.io/dxwg/dcat/#qualified-attribution.

Cheers,

Nick
(a provenance researcher in Australia)


From: Matthew Brush <brushm@ohsu.edu<mailto:brushm@ohsu.edu>>
Date: Wednesday, 25 September 2019 at 5:50 pm
To: "public-prov-comments@w3.org<mailto:public-prov-comments@w3.org>" <public-prov-comments@w3.org<mailto:public-prov-comments@w3.org>>
Subject: PROV use for a Contributor Attribution Model


Hello, I wanted to share a modeling specification for representing contributions to research-related artifacts called the Contributor Attribution Model (CAM), which partially overlaps in scope with PROV. It is still an early effort, and developing documentation is available here [1]. I will briefly summarize some context relevant to my question below, about the possibility of PROV alignment and re-use.



The scope of the CAM data model is specifically limited to representing the nature of an agent’s contribution to a research-related artifact.  It is meant to be used as a module within a larger data model that captures the complete semantics of a given domain or use case.  Its driving use cases require as *simple* and *direct* a model as possible for representing contribution metadata.  Accordingly, we define a three-object structure in which a Contribution object mediates the link between an Entity (aka Artifact) and a contributing Agent.  The Contribution object holds attributes to describe when, where, how, and in what context the contribution was made.  See here [2].



We of course explored how we might use PROV to implement this structure, but ran up against an issue that I have seen previously debated in this forum (see [3], [4], [5], [6]) concerning restrictions on prov:Attribution objects - which are analogous to cam:Contributions in that they link an Entity to a contributing Agent.  Specifically, PROV does not permit things like Roles, Plans, and Locations to hang from this class of objects.  Rather, they are permitted only on qualified relations involving an Activity (e.g. an Association, Generation, or Usage).  Because of this, use of PROV to address our needs would require a four object model that includes an Entity, Activity, Association, and Agent.



This additional complexity was one of the main reasons we decided to develop the CAM model. We will of course provide mappings to PROV, but we would like to explore if/how we might use PROV directly to build the simpler structure we need.  I was wondering if, given the history of questions and requests concerning the aforementioned restrictions on Attributions, we might revisit if/how these restrictions might be relaxed in a way that mitigates the concerns that motivated them in the first place - specifically, ambiguity about whether the Artifact or Agent participating in a qualified Attribution bears the role.  Alternatively, if this ship has sailed, perhaps there is a mechanism that would allow us to define a PROV extension or profile that meets our needs in a PROV compliant way.



Thanks for your consideration, and we look forward to exploring alignment opportunities.

 Matthew Brush



[1]CAM ReadtheDocs Home:  https://contributor-attribution-model.readthedocs.io/en/latest/index.html

[2] CAM Information Model:  https://contributor-attribution-model.readthedocs.io/en/latest/info_model/index.html

[3] Question about attaching Roles to Attribution objects: https://lists.w3.org/Archives/Public/public-prov-comments/2018Jun/0004.html

[4] Another question related to this: https://www.w3.org/2011/prov/track/issues/384

[5] 'Resolution' on the role question: https://www.w3.org/2011/prov/wiki/ResponsesToPublicComments#ISSUE-532_.28Role.29

[6] Time/Location attributes on Attribution objects: https://www.w3.org/2011/prov/wiki/index.php?title=ResponsesToPublicComments&mobileaction=toggle_view_desktop#ISSUE-530_.28attributes.29


-------------------------
Matthew H. Brush, PhD
Research Assistant Professor
Department of Medical Informatics and Clinical Epidemiology
Translational & Integrative Sciences Lab
Oregon Health and Science University
Portland, OR, USA
brushm@ohsu.edu<mailto:brushm@ohsu.edu>