[poe] ODRL Ontology: formal definition and design issues

nitmws has just created a new issue for https://github.com/w3c/poe:

== ODRL Ontology: formal definition and design issues ==
Having a deeper look at the whole ODRL Ontology - http://w3c.github.io/poe/vocab/ODRL22.ttl as of 2 June - I noticed these issues:

1. Formal use of [SKOS ](https://www.w3.org/TR/skos-reference/)
a) The atomic particles of SKOS are skos:Concepts and the containers skos:ConceptScheme or skos:Collection can include skos:Concepts only. Only the skos:Collection's #actions and #actionsCommon include skos:Concepts, all others don't include any - _this is a bug_.
b) Use of skos:ConceptScheme (currently only actions are a Concept Scheme)
b.1) ... requires that all concepts assigned to a scheme indicate that by using skos:inSchema. None of the actions does that - - _this is a bug_.
b.2) the scheme :actions lists multiple concepts as skos:hasTopConcept. This is only right for :use and :transfer, all other concepts are narrower than one of these two and therefore not a TopConcept - _this is a bug_

2. hasPolicy property: a policy does NOT have a "target Asset" therefore the _definition is wrong_ - and the rdfs:range :policy needs a better definition. (E.g. that the related Policy includes at least one Permission or Prohibition with this Asset as target.)

3. using both rdfs:Class and owl:Class as type. http://www.w3.org/TR/owl-rdf-based-semantics defines
rdfs:Class rdfs:subClassOf owl:Class .
My view: rdfs:Class only is sufficient.

4. using both rdf:Property and owl:ObjectProperty: http://www.w3.org/TR/owl-rdf-based-semantics defines
rdf:Property rdfs:subClassOf owl:ObjectProperty .
My view: rdf:Property only is sufficient.

5. Design of defining individuals
Note: the SKOS specification makes a statement about "SKOS Concept Schemes and OWL Ontologies" - https://www.w3.org/TR/skos-reference/#L1170 - declaring a graph can be both, a SKOS Concept Scheme and an OWL Ontology. 
The SKOS Recommendation explains that SKOS Concept individuals are equivalent to OWL individuals - https://www.w3.org/TR/skos-reference/#L1045 
a) Why are only the ODRL-defined individuals of the Action class a skos:Concept and the individuals of the LeftOperand and Operator class are owl:NamedIndividuals? They all have the same role as individuals of a class. I propose to harmonise them - preferred: making them skos:Concepts.
b) Currently ODRL defines for all individuals of Action a single Concept Scheme with the identifier http://www.w3.org/ns/odrl/2/actions. The normative and non-normative concepts are sorted out only by SKOS Collections with the IRIs http://www.w3.org/ns/odrl/2/#actions and http://www.w3.org/ns/odrl/2/#actionsCommon - these IRIs are very close to the scheme IRI. I propose to split them into 2 schemes: http://www.w3.org/ns/odrl/2/actionsCommon and to drop the use of Collections.
c) For what purpose are skos:Collections used beyond the purpose above? If they should show the "building blocks" of a class they are not complete, e.g. http://www.w3.org/ns/odrl/2/#ruleConcepts includes only :Rule, :relation and :function - but no :permission, :prohibition or :constraint. And as pointed out above: the class and the properties are formally invalid in a SKOS Collection.





Please view or discuss this issue at https://github.com/w3c/poe/issues/188 using your GitHub account

Received on Sunday, 4 June 2017 17:55:00 UTC