RE: Another approach to the non-reification problem...

Thanks for this Ivan. This option did cross my mind but you’ve brought it into focus.

Named graphs often provide elegant solutions to these kinds of problems. They are certainly our favourite approach to issues of provenance, rights, and some temporal information.

However I would hesitate to use them as a general modeling tool. We’re only just learning how to manage named graphs: when to create them; how to split them. I’d be wary of putting more weight on them.

What do you think Renato? Do others in the group have experience of using named graphs at scale?


From: Ivan Herman [mailto:ivan@w3.org]
Sent: 26 September 2016 10:04
To: Dr. Renato Iannella; Whittam Smith, Benedict (TR Technology & Ops)
Cc: W3C POE WG
Subject: Another approach to the non-reification problem...

Hi guys,

I just saw a relatively similar question coming up on the SWIG mailing list. The solution put forward was to use named graphs. (I believe NG-s were not yet fully finalized as standards when the the Prov WG was working.)

To make it a bit abstract (Ben, you tell me if this understanding of the problem is what we are looking for.) We have a triple:

          ex:A ex:B ex:C.

Where ex:B is of class ex:Prop

And we want, somehow to attach additional information on the triple itself (eg, say that the triple's permissible time interval is such and such).

1. The Prov solution:

defined a 'qualified class', ie, ex:qualifiedProp.

ex:B
   rdf:type    ex:qualifiedProp;
   ex:source ex:A;
   ex:target  ex:C;
   … add any additional triples here

2. The NG solution:

   ex:mySpecificTriple { ex:A ex:B ex:C }.
   ex:mySpecificTriple
      … add any additional triples here

Overall:

- both are convoluted
- both are unfriendly for JSON-LD, but both can be encoded.
- the NG looks nicer in turtle, oops, not, it is then TriG (and most of the systems out there understand NG-s)
- conceptually, if NG-s are used, a full ODRL is not an RDF Graph, but an RDF Dataset[1]
- both represent, I presume, an extra complication for the semantics document, and I am not sure whether there are OWL formalizations of datasets (the formal specification of Datasets came after the publication of OWL2)
- NG-s are somewhat cleaner and more in line with the RDF standards...

Hope this helps!

Cheers

Ivan


[1] https://www.w3.org/TR/rdf11-concepts/#section-dataset


----
Ivan Herman, W3C
Digital Publishing Lead
Home: http://www.w3.org/People/Ivan/

mobile: +31-641044153
ORCID ID: http://orcid.org/0000-0003-0782-2704