W3C home > Mailing lists > Public > public-dwbp-wg@w3.org > December 2015

Re: Removed domain and range from 3rd party vocabularies

From: Eric Stephan <ericphb@gmail.com>
Date: Sat, 19 Dec 2015 09:21:01 -0800
Message-ID: <CAMFz4jiVjEyPSLY=bKzouvNHAhZaTWxABW_ZB6KnR4Ue8hvdXA@mail.gmail.com>
To: Laufer <laufer@globo.com>, Bernadette Farias Loscio <bfl@cin.ufpe.br>, "Purohit, Sumit" <sumit.purohit@pnnl.gov>
Cc: Public DWBP WG <public-dwbp-wg@w3.org>
Laufer,

Thank you once again for reviewing and providing invaluable feedback.  I
changed the http://w3c.github.io/dwbp/vocab-du.html to fix the isCitedBy
discrepancy.

I am thinking about what you are stating with regards to inferencing.  I
found this concrete example and the problem of inferencing with multiple
domain declarations.
http://stackoverflow.com/questions/30172835/multiple-domain-and-range-in-objectproperty
.  This could be a limiting factor or headache for someone in the future.
Just an observation about vocabulary re-use.  If I were to be designing a
collection of vocabularies from the top down, domains and ranges make a lot
of sense.  But if you are just providing a domain and range to a vocabulary
that may be used anywhere, its  more important to convey the usage
information to the human (knowledge worker) assembling the pieces together
for a particular application.

Berna, Sumit what do you think?

Also just to document as well, At one point we had prov:Entity for the
dcat:Dataset and dcat:Distribution.  After speaking with Antoine I am
convinced that having a superclass without any purpose than creating a
level of abstraction will probably get in the way of most people using the
vocabulary.


On Sat, Dec 19, 2015 at 8:49 AM, Laufer <laufer@globo.com> wrote:

>
> Hi Eric,
>
> First of all, thank you for your efforts in writing the document.
>
> My concern is about the formal definition of duv. This formal definition
> could be used by reasoners to conclude things.
>
> I liked the use of vann:Usage.
>
> I still have some concerns:
>
> 1. In the diagram there is a property "cito:isCitedBy" that appears in the
> specification as "duv:isCitedBy";
>
> 2. The range of the property "duv:refersTo" has two possible classes:
> "dcat:Dataset" and "dcat:Distribution". If that way, when someone defines
> an usage, an inference machine will conclude that it refers to a resource
> that is both a "dcat:Dataset" and a "dcat:Distribution". And I think this
> is not the case.
>
>   I see three options:
>     a) one single property: "duv:refersTo" without range definitions;
>     b) two properties: "duv:refersToDataset" with range "dcat:Dataset",
> and "duv:refersToDistribution" with range "dcat:Distribution";
>     c) options a) and b) with properties "duv:refersToDataset"
> and "duv:refersToDistribution" as subproperties of  "duv:refersTo" (I
> prefer this).
>
> 3. The same thing occurs in respect to the domain of properties
> "duv:hasUsage" and "duv:hasUserFeedback".
>
> Thank you a lot again.
>
> Cheers, Laufer
>
> --
>
> .  .  .  .. .  .
> .        .   . ..
> .     ..       .
>
>
>
> Em 19/12/2015 12:23, Eric Stephan escreveu:
>
> Hi Laufer,
>
> I've removed the domain and range from 3rd party properties and replaced
> them with a vann:Usage statement as Phil suggested to at least document the
> preferred use of the vocabulary from the DUV perspective.
>
> Could you take a look when you get a chance to see if this satisfies your
> concerns?
>
> Thanks,
>
> Eric S
>
>
Received on Saturday, 19 December 2015 17:21:30 UTC

This archive was generated by hypermail 2.3.1 : Saturday, 19 December 2015 17:21:31 UTC