Re: "globally closed" schema?

Hello

On a (maybe) related note : I have a use-case that involves documenting a
dataset (not necessarily validating it) against a SHACL specification. By
documenting I mean provide a "summary" of the graph.
To do this I would like to *count* the number of targets of each shape in
the dataset; for this :

   1. I could use a more direct predicate to indicate a node is a target of
   a shape (e.g. "xx:isTargetOfShape") (which would then be the basis to
   determine if a node is not targeted by any shape). This could be seen as
   the inverse of sh:targetNode.
   2. I would like to express the statistics of the dataset against the
   SHACL specification, that is the number of targets of each node shape, the
   number of occurrences of each property shape, and the number of distinct
   values of each property shape. For this I can rely on VOID vocabulary (
   https://www.w3.org/TR/void/#statistics), but only if the node shapes
   corresponds to classes (i.e. use sh:targetClass) to use
   void:classPartition. To generalize this partitioning approach (
   https://www.w3.org/TR/void/#subset) I could also use a more
   SHACL-related partitioning of a dataset, such as xx:shapePartition, with a
   "xx:shape" property pointing to a shape. This partition would "contain all
   triples that describe entities that are targets of the shape indicated with
   xx:shape." The statistics are then expressed on the partition entity.

Best Regards
Thomas



Le jeu. 18 janv. 2024 à 18:44, Holger Knublauch <holger@topquadrant.com> a
écrit :

>
>
> On 18 Jan 2024, at 4:29 pm, Vladimir Alexiev <
> vladimir.alexiev@ontotext.com> wrote:
>
> sh:closed is *local*, i.e. per shape. It checks that a *targeted *node
> doesn't include *unsanctioned props*.
> I'm looking for ways to declare a *globally *closed schema, i.e. to catch *untargeted
> *nodes, i.e. that the KG doesn't include *unsanctioned nodes*.
>
> 1. I guess one can do it with SPARQL (SHACL advanced) by triggering off
> some fixed node.
>   Here's a very expensive way to catch all nodes: select distinct ?s {?s
> ?p ?o}
> 2. It would be useful to "mark" all nodes visited during SHACL validation.
> - SHEX has a way to report positive results:
> https://shexspec.github.io/primer/ShExJ#validation
> - SHACL doesn't have a standard way to report positive results:
> https://www.w3.org/TR/shacl/#results-validation-result .
>   But there is a note "SHACL implementations may use other SHACL
> subclasses of sh:AbstractResult, for example, to report successfully
> completed constraint checks or accumulated results."
>   Say we "standardize" a new class eg dash:PositiveValidation
>
>
> Already exist:
>
> dash:SuccessResult
>   a rdfs:Class ;
>   rdfs:comment "A result representing a successfully validated constraint." ;
>   rdfs:label "Success result" ;
>   rdfs:subClassOf sh:AbstractResult ;
> .
>
>
> Then one could look for nodes that don't appear in
> sh:ValidationResult.focusNode nor dash:PositiveValidation.focusNode
>
>
> Yes.
>
> There may also be different algorithms such as enumerating the allowed
> rdf:types which could be represented with a targeted constraint on rdf:type
> and a sh:in. This depends on what someone considers unsanctioned nodes.
>
> Holger
>
>
>
> All feedback is welcome!
>
>
>

-- 

*Thomas Francart* -* SPARNA*
Web de *données* | Architecture de l'*information* | Accès aux
*connaissances*
blog : blog.sparna.fr, site : sparna.fr, linkedin :
fr.linkedin.com/in/thomasfrancart
tel :  +33 (0)6.71.11.25.97, skype : francartthomas