owl-shacl question

[Bohms, H.M. (Michel)]

Dear Irene, Holger

 

In http://spinrdf.org/shacl-and-owl.html it is suggested to have the following ‘architecture’:

 

 

Would I lose much benefit when I simplify a bit by just having OWL/RDFS (with owl/rdfs owa inferencing) on top and (several) SHACL views (with shacl cwa validation and inferencing (rule part)) below. (I just mention ‘several’because typically there are more validation-views possible working on the same conceptual owl-view). Just looking for pros and cons. One of the reasons is that in practically all existing ontologies relevant for us do not have the split yet in rdfs and owl…just an example: a very recent proposal that I really like: https://www.w3.org/TR/owl-time/, is in OWL-DL. I want to directly reuse and pref. not split first in a RDFS part and an OWL-DL part.

 

Thx for your advice here,

Michel

 

Dr. ir. H.M. (Michel) Böhms Senior Data Scientist T +31888663107 M +31630381220 E michel...@tno.nl Location   This message may contain information that is not intended for you. If you are not the addressee or if this message was sent to you by mistake, you are requested to inform the sender and delete the message. TNO accepts no liability for the content of this e-mail, for the manner in which you use it and for damage of any kind resulting from the risks inherent to the electronic transmission of messages.

 

 

[Holger Knublauch]

Sure, that should work without problems. The suggested split is mostly for new projects that are under your control.

Holger

Sent from my iPad

On 4 Oct 2017, at 16:49, Bohms, H.M. (Michel) <michel...@tno.nl> wrote:

Dear Irene, Holger

 

In http://spinrdf.org/shacl-and-owl.html it is suggested to have the following ‘architecture’:

 

<image002.png>

 

Would I lose much benefit when I simplify a bit by just having OWL/RDFS (with owl/rdfs owa inferencing) on top and (several) SHACL views (with shacl cwa validation and inferencing (rule part)) below. (I just mention ‘several’because typically there are more validation-views possible working on the same conceptual owl-view). Just looking for pros and cons. One of the reasons is that in practically all existing ontologies relevant for us do not have the split yet in rdfs and owl…just an example: a very recent proposal that I really like: https://www.w3.org/TR/owl-time/, is in OWL-DL. I want to directly reuse and pref. not split first in a RDFS part and an OWL-DL part.

 

Thx for your advice here,

Michel

 

Dr. ir. H.M. (Michel) Böhms Senior Data Scientist T +31888663107 M +31630381220 E michel...@tno.nl Location

<image003.gif>

This message may contain information that is not intended for you. If you are not the addressee or if this message was sent to you by mistake, you are requested to inform the sender and delete the message. TNO accepts no liability for the content of this e-mail, for the manner in which you use it and for damage of any kind resulting from the risks inherent to the electronic transmission of messages.

 

 

--
You received this message because you are subscribed to the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to topbraid-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Bohms, H.M. (Michel)]

Ok, any pros for the new approach (compared to just owl-rdfs/shacl split) very welcome,

Gr Michel

 

 

 

Dr. ir. H.M. (Michel) Böhms Senior Data Scientist T +31888663107 M +31630381220 E michel...@tno.nl Location

[Holger Knublauch]

The pros are that users that don't want/need to use OWL don't have to import these axioms, and thus the RDF files that you are producing are more focused - think of separation of concerns, which is generally a good engineering practice. SHACL doesn't need OWL, OWL doesn't need SHACL, so there is no reason to stack them on top of each other. If projects plan ahead they can also more easily retire any part of their stack that is no longer needed. Finally, having OWL axioms around may create false expectations or confuse the view point from a SHACL perspective (although the axioms are technically ignored, users may expect inferencing to happen).

Holger

[Bohms, H.M. (Michel)]

Thx Holger, for the argumentation (SoC)!

 

Just….

I could see a scenario in future where in CWA-area OWL could be fully replaced by SHACL (ie retire OWL). Because there are no dependencies as you explain below there are certainly no technical restraints for something like that.

But then…..aren’t we losing more that we want? Clearly we want to replace OWA Restrictions to SHACL Shapes. But what about eg the distinction in datatype properties and object properties and other very useful OWL modelling constructs…ie wouldn’t there be a need for an OWL Light or RDFS+ to combine cleanly with SHACL then?

 

Thanks very much for your views again,

[Holger Knublauch]

See https://twitter.com/wohnjalker/status/915982539747028992

As a slightly more serious response, I agree that URIs from the OWL namespace may be useful even without OWL semantics. owl:imports is clearly useful, and even referenced by the SHACL spec. owl:versionInfo and the deprecation mechanisms can be useful, but they don't carry OWL semantics. Whether owl:DatatypeProperty and owl:ObjectProperty provide value is a matter of debate. I believe as long as there are sh:class and sh:datatype or sh:nodeKind constraints in place, then there is no need for them. I am not fond of global property axioms in general, but that's another topic.

Maybe there is value in going through the ways that people have used OWL so far and verify how many of them were really designed for OWL (DL) inferencing. Maybe you have examples of axioms in your world, that you could share here so that we can see what would be left that isn't covered by SHACL or other non-OWL vocabularies.

Holger

[Bohms, H.M. (Michel)]

 

 

See https://twitter.com/wohnjalker/status/915982539747028992

Ø    Great summary! 😊

As a slightly more serious response, I agree that URIs from the OWL namespace may be useful even without OWL semantics. owl:imports is clearly useful, and even referenced by the SHACL spec. owl:versionInfo and the deprecation mechanisms can be useful, but they don't carry OWL semantics. Whether owl:DatatypeProperty and owl:ObjectProperty provide value is a matter of debate. I believe as long as there are sh:class and sh:datatype or sh:nodeKind constraints in place, then there is no need for them. I am not fond of global property axioms in general, but that's another topic.

Maybe there is value in going through the ways that people have used OWL so far and verify how many of them were really designed for OWL (DL) inferencing. Maybe you have examples of axioms in your world, that you could share here so that we can see what would be left that isn't covered by SHACL or other non-OWL vocabularies.

>well so far the distinction between attributes/datatypeproperties and relationships/objectproperties has proven useful since they in the end say something about intrinsic properties of things and the more role-based extrinsic properties towards other independent things which is a quite basic notion in conceptual modelling not only in LD/SW but in any other earlier modelling system. But always interesting of course to rethink….

> owl:import hopefully obsolete in a future where all is dereferenceable…

>owl:equivalent could be two way rdfs:subClassOf of course

>owl:sameAs is seen as important but has big issues too (is it really sameAs that you want etc. ie what does it mean, don’t you really want a weaker thing; and in CWA can be done via UNA anyway)

> unionOf/intersectionOf but I expect they have counterpart in shacl

> inverse properties

>disjointWith /propertydisjointWith

> So actually main concern is distinction in attributes and relationships (or if you like values and references) in the end…. 😊

Gr michel

[Holger Knublauch]

On 6/10/2017 18:44, Bohms, H.M. (Michel) wrote:

 

 

See https://twitter.com/wohnjalker/status/915982539747028992

Ø    Great summary! 😊

As a slightly more serious response, I agree that URIs from the OWL namespace may be useful even without OWL semantics. owl:imports is clearly useful, and even referenced by the SHACL spec. owl:versionInfo and the deprecation mechanisms can be useful, but they don't carry OWL semantics. Whether owl:DatatypeProperty and owl:ObjectProperty provide value is a matter of debate. I believe as long as there are sh:class and sh:datatype or sh:nodeKind constraints in place, then there is no need for them. I am not fond of global property axioms in general, but that's another topic.

Maybe there is value in going through the ways that people have used OWL so far and verify how many of them were really designed for OWL (DL) inferencing. Maybe you have examples of axioms in your world, that you could share here so that we can see what would be left that isn't covered by SHACL or other non-OWL vocabularies.

>well so far the distinction between attributes/datatypeproperties and relationships/objectproperties has proven useful since they in the end say something about intrinsic properties of things and the more role-based extrinsic properties towards other independent things which is a quite basic notion in conceptual modelling not only in LD/SW but in any other earlier modelling system. But always interesting of course to rethink….

> owl:import hopefully obsolete in a future where all is dereferenceable…

>owl:equivalent could be two way rdfs:subClassOf of course

>owl:sameAs is seen as important but has big issues too (is it really sameAs that you want etc. ie what does it mean, don’t you really want a weaker thing; and in CWA can be done via UNA anyway)

> unionOf/intersectionOf but I expect they have counterpart in shacl

> inverse properties

>disjointWith /propertydisjointWith

> So actually main concern is distinction in attributes and relationships (or if you like values and references) in the end…. 😊

On the owl:Datatype/ObjectProperty topic, why would sh:datatype/sh:class/sh:nodeKind not be sufficient? The only difference is that the latter are per-shape, which is similar to how most other languages such as UML or XML handle these things - the concept of global property declarations are pretty unique to semantic web technology. But for a tool there is little difference on whether you check (in SPARQL)

?property a owl:DatatypeProperty

vs

?ps sh:path ?property .
?ps sh:nodeKind sh:Literal .

Note that even the former case (in OWL) is not sufficient - you would still need to add code to query for the cases of untyped properties or properties that have rdf:type rdf:Property and then declare an rdfs:range or owl:allValuesFrom restrictions. The SHACL variant looks more pragmatic compared to that.

Holger

[Bohms, H.M. (Michel)]

Ok clear...so owl only offers in the end some more (global) semantic sugar here....I assume the uncontrained property sec IS still globally available as rdf:property, right? (and just locally/atclass RESTRICTED)
Thx Michel

Verzonden van mijn Android-telefoon via TouchDown (www.symantec.com)

-----Original Message-----
From: Holger Knublauch [hol...@topquadrant.com]
Received: zaterdag, 07 okt. 2017, 1:15
To: topbrai...@googlegroups.com [topbrai...@googlegroups.com]
Subject: Re: [topbraid-users] owl-shacl question

--
You received this message because you are subscribed to the Google Groups "TopBraid Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to topbraid-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

 

This message may contain information that is not intended for you. If you are not the addressee or if this message was sent to you by mistake, you are requested to inform the sender and delete the message. TNO accepts no liability for the content of this e-mail, for the manner in which you use it and for damage of any kind resulting from the risks inherent to the electronic transmission of messages.

[Holger Knublauch]

On 8/10/2017 1:59, Bohms, H.M. (Michel) wrote:

Ok clear...so owl only offers in the end some more (global) semantic sugar here....I assume the uncontrained property sec IS still globally available as rdf:property, right? (and just locally/atclass RESTRICTED)

Yes, defining an rdf:Property is part of RDF/Schema. Also declaring rdfs:label and rdfs:comment. SHACL has per-shape equivalents for these (sh:name, sh:description).

Holger

[Bohms, H.M. (Michel)]

 

Dear Holger,

One more issue on OWL versus SHACL.

Wrt owl:sameAs its says at http://spinrdf.org/shacl-and-owl.html:

 

So it says: no need because diff. by default.

 

I can understand this in case of a CWA situation. But what about in a link between two resources from two different/independent parties (say both CWA).

 

I would still need owl:sameAs to connect those right?

 

(it feels a bit like global OWA/local CWA: anybody can make his own CWA ontology).

 

In case the two parties would agree some reference data set (in my case the names/ids for NL roads) I guess this agreement creates a kind of common CWA situation between the two parties and then again indeed owl:sameAs is not needed anymore (?).

 

Thanks for your view here,

Michel

 

 

Dr. ir. H.M. (Michel) Böhms Senior Data Scientist T +31888663107 M +31630381220 E michel...@tno.nl Location

This message may contain information that is not intended for you. If you are not the addressee or if this message was sent to you by mistake, you are requested to inform the sender and delete the message. TNO accepts no liability for the content of this e-mail, for the manner in which you use it and for damage of any kind resulting from the risks inherent to the electronic transmission of messages.

[Irene Polikoff]

Michel,

You can use owl:sameAs between resources. SHACL will not do any reasoning based on this statement.

As you mentioned, the semantics of owl:sameAs is somewhat problematic. In most cases, users do not want or expect sameAs inferencing as intended by OWL. They typically want some meaning and conclusions specific to their goals. We often recommend the use of skos:exactMatch or some other mapping/matching property including one(s )from your own vocabulary for which you can define your own specific semantics.

On Oct 9, 2017, at 1:59 PM, Bohms, H.M. (Michel) <michel...@tno.nl> wrote:

 

Dear Holger,

One more issue on OWL versus SHACL.

Wrt owl:sameAs its says at http://spinrdf.org/shacl-and-owl.html:

 

<image001.png>

So it says: no need because diff. by default.

 

I can understand this in case of a CWA situation. But what about in a link between two resources from two different/independent parties (say both CWA).

 

I would still need owl:sameAs to connect those right?

 

(it feels a bit like global OWA/local CWA: anybody can make his own CWA ontology).

 

In case the two parties would agree some reference data set (in my case the names/ids for NL roads) I guess this agreement creates a kind of common CWA situation between the two parties and then again indeed owl:sameAs is not needed anymore (?).

 

Thanks for your view here,

Michel

Dr. ir. H.M. (Michel) Böhms Senior Data Scientist T +31888663107 M +31630381220 E michel...@tno.nl Location

<image002.gif>