Yes, indeed this evaluation is part of my thesis :)
Each of these requirements to formulate constraints and to validate RDF data against these constraints corresponds to a type of constraints against which RDF data is to be validated.
When I determined this set of 81 requirements / constraint types, this set encompassed the union of
1. constraint types which are supported by current high-level constraint languages (ShEx, ReSh, DSP, OWL)
2. requirements identified and (not yet) approved by the W3C WG
3. requirements identified and (not yet) approved by the DCMI WG
As these requirements are continuously updated, the wording may have changed for some of them.
As to a certain time I had to make a cut, there are certainly newly defined requirements now which are not part of this set of 81 requirements.
I implemented these constraint types when they are expressible in OWL and in DSP [1] which can be tested in the validation environment [2].
Kai and I defined a simple vocabulary which can be used to formulate constraints of each constraint type in a concise way.
I attach the paper Kai presented at the Dublin Core conference this year.
When I use OWL for RDF validation I assume CWA and UNA.
I attachted a paper (for the SEMANTiCS conference this year)
where together with Kai we investigated
for each constraint type
(1) if reasoning may be performed prior to validation to enhance data quality and
(2) if validation results differ when different semantics is assumed.
Thomas
[1] https://github.com/github-thomas-hartmann/phd-thesis/tree/master/rdf-validation/constraint-languages/implementations
[2] http://purl.org/net/rdfval-demo
Hi Thomas,
Thanks a lot!
This is really interesting, and frankly impressive. I guess this comes fresh out of a thesis ;-)
I have a some very general questions:
- where are the 81 'validation requirements' coming from? Are they a superset of the DC ones [1]? Or the W3C ones? If it's W3C then SHACL seems to be missing a lot...
- are the results based on desk research or do you have a suite of 81 formalized test cases and their implementation in all these languages?
- when you mention OWL, is it wrt the 'traditional' OWL semantics or are you considering the 'closed world hijacking' of it?
Cheers,
Antoine
[1] http://wiki.dublincore.org/index.php/RDF_Application_Profiles/Requirements
On 12/3/15 1:55 PM, Antoine Isaac wrote:
> [I'm reposting Thomas' mail in its own thread. I think it deserves it! - Antoine]
>
> -------- Original Message --------
> From: Hartmann, Thomas <[log in to unmask]>
>
> Hi all,
>
> I conducted an evaluation to what extent the constraint languages SHACL, ShEx, ReSh, OWL 2, DSP, and SPIN cover RDF validation requirements.
> Attached you find the evaluation results.
>
> Many / Most of the DC requirements are included in this evaluation (although the wording may be different in some cases, since our database is continuously updated),
> but not all - i.a., the ones not approved by the W3C working group.
>
> I could walk through our DC requirements and continue this evaluation for the remaining DC requirements, if people think this is useful.
>
>
> Cheers,
> Thomas
>
> --
> Thomas Hartmann (formerly Bosch)
> M.Sc. in Information Systems, Ph.D. Student in Computer Science
>
> GESIS - Leibniz Institute for the Social Sciences
> Social Science Metadata Standards
> Visitors Address: B2,1, D-68159 Mannheim
> Tel: + 49 (0) 621 / 1246-271
> Web: http://boschthomas.blogspot.com/
|