- From: Arthur Ryman <arthur.ryman@gmail.com>
- Date: Thu, 5 Nov 2015 11:50:42 -0500
- To: Holger Knublauch <holger@topquadrant.com>
- Cc: "public-data-shapes-wg@w3.org" <public-data-shapes-wg@w3.org>
All, Good discussion. Holger only voted -0.5 on the straw poll and seems willing to look at a concrete example vocabulary if someone produces one. I will take that as an action for SHACL. fyi, Here is the last vocabulary I wrote at IBM in Turtle [1] and HTML [2]. These are served up with content negotiation when you dereference the vocabulary IRI http://jazz.net/ns/process#. The Turtle file is hand-written source. The HTML is generated from the Turtle using Jena to convert the Turtle to non-abbreviated RDF/XML, and then an XSLT to transform the RDF/XML to HTML. If we like this general approach then I would re-implement it to fit in with ReSpec. Probably convert the Turtle to JSON-LD and then generate the HTML using JavaScript in a ReSpec plugin. [1] https://jazz.net/wiki/pub/LinkedData/JazzProcessVocabulary/process-vocabulary.ttl [2] https://jazz.net/wiki/bin/view/LinkedData/JazzProcessVocabulary -- Arthur On Wed, Nov 4, 2015 at 6:06 PM, Holger Knublauch <holger@topquadrant.com> wrote: > FYI I have contacted (my former colleague) Matthew Horridge on the OWL > API/Protege issues and will report back on his findings. As these issues > appear to be general problems with the handling of RDFS files, I hope they > can be addressed generically by their implementation. Regardless, we may > still consider publishing an (alternative) OWL version of the SHACL > namespace, to optimize the interaction with such tools. We all want to lower > the barriers of adoption. > > Holger > > > > On 11/4/2015 6:30, Peter F. Patel-Schneider wrote: >> >> I agree that it is not exactly an OWL problem, but several OWL tools are >> quite >> expansive in what they can reasonably handle and it seems to me that >> property >> punning should be easy to handle even without typing statements. I'm >> trying >> to see what can be done to improve the tools involved. >> >> I also do not understand the resistance to RDFS so I can't help you there. >> >> peter >> >> >> On 11/03/2015 12:11 PM, Karen Coyle wrote: >>> >>> Peter, I actually do not see this as purely a OWL problem, nor do I think >>> that >>> the tools that function well with RDFS and OWL need to change. It would >>> be >>> convenient in many circumstances to have ranges, and perhaps also >>> domains, >>> defined for the properties in SHACL, since both humans and software do >>> frequently make use of that information. Perhaps someone can explain the >>> resistance against this? >>> >>> kc >>> >>> On 11/3/15 9:04 AM, Peter F. Patel-Schneider wrote: >>>> >>>> If you ask for >>>> >>>> https://github.com/TopQuadrant/shacl/blob/master/src/main/resources/etc/shacl.ttl >>>> >>>> without content negotiation you will get back an HTML document. If you >>>> ask >>>> for this document in text/turtle, you get back a 406 error. So it is >>>> not >>>> surprising that when you try to load the document into Protege there are >>>> complaints. >>>> >>>> If you ask for >>>> >>>> https://raw.githubusercontent.com/TopQuadrant/shacl/master/src/main/resources/etc/shacl.ttl >>>> >>>> you get something that can be loaded into Protege. However, there >>>> appears to >>>> be some problem with whatever is converting the document into OWL and >>>> the >>>> properties that are also individuals are not being handled nicely. >>>> >>>> Note that the document is not an OWL 2 DL document so the tools that are >>>> loading it are trying to determine what changes are needed to turn it >>>> into OWL >>>> 2 DL. It appears that one of these tools isn't doing something that it >>>> seems >>>> would be quite easy. >>>> >>>> peter >>>> >>>> >>>> On 11/03/2015 07:47 AM, Karen Coyle wrote: >>>>> >>>>> I am able to load the file into Protege, but not directly from the IRI >>>>> (which >>>>> gives an error - probably some github thing). I can load a saved >>>>> version. >>>>> However, that saved version appears to lack all of the properties that >>>>> are >>>>> listed at the bottom of the ttl file -- and, of course, those are >>>>> exactly what >>>>> I am hoping to work with to create a SHACL document. >>>>> >>>>> I expected those properties to be treated as OWL annotation properties >>>>> because >>>>> they do not have domains or ranges defined. (That is what happens to >>>>> Dublin >>>>> Core 1.1 terms.) Instead, they are simply missing. This is clearly a >>>>> "feature" >>>>> of the Manchester API, but it seems to be one that could easily be >>>>> avoided by >>>>> adding ranges to those properties. >>>>> >>>>> kc >>>>> >>>>> On 11/2/15 10:10 PM, Holger Knublauch wrote: >>>>>> >>>>>> I had tried my latest version from >>>>>> >>>>>> >>>>>> https://github.com/TopQuadrant/shacl/blob/master/src/main/resources/etc/shacl.ttl >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> Holger >>>>>> >>>>>> >>>>>> On 11/3/2015 15:24, Peter F. Patel-Schneider wrote: >>>>>>> >>>>>>> I'm not sure what file you are talking about? >>>>>>> >>>>>>> The SHACL Vocabulary description linked to from the WG Wiki home page >>>>>>> is >>>>>>> >>>>>>> https://raw.githubusercontent.com/w3c/data-shapes/56429ef268a14e29586244ab944b310ea84cbf46/shacl/shacl.shacl.ttl >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> The file I get from this location does not have the string >>>>>>> "AbstractResult" in >>>>>>> it at all. >>>>>>> >>>>>>> peter >>>>>>> >>>>>>> >>>>>>> On 11/02/2015 08:35 PM, Holger Knublauch wrote: >>>>>>>> >>>>>>>> On 11/3/2015 14:04, Peter F. Patel-Schneider wrote: >>>>>>>>> >>>>>>>>> I just loaded the file pointed at from the WG home page into >>>>>>>>> Protege >>>>>>>>> 5.0.0. >>>>>>>>> No problems loading or classifying, but of course there isn't much >>>>>>>>> that an OWL >>>>>>>>> reasoner does with it. >>>>>>>>> >>>>>>>>> What part of RDF does the OWL API not support? >>>>>>>> >>>>>>>> To see some of the problems, re-save the shacl file with Protege. >>>>>>>> Then compare >>>>>>>> it with the original triples. There are all kinds of changes, e.g. >>>>>>>> rdfs:Classes suddenly have become owl:Classes and >>>>>>>> owl:NamedIndividual. Some >>>>>>>> triples seem to have just disappeared (e.g. >>>>>>>> sh:optionalWhenInherited, >>>>>>>> sh:maxCount at sh:AbstractResult, the sh:constraint at sh:Argument). >>>>>>>> >>>>>>>> See the attached screenshot of an example diff. >>>>>>>> >>>>>>>> HTH >>>>>>>> Holger >>>>>>>> >>>>>> >>>>>> > >
Received on Thursday, 5 November 2015 16:51:11 UTC