Modeling support for "Negative Type" statement about owlsharp

[mdesalvo]

Hi @constnick,
I added two new methods in the Data object to fulfill the new use case:

• DeclareNegativeIndividualType: to explicitly state that the given individual is not of the given class

OWLOntology ontology = new OWLOntology("ex:ont");
ontology.Model.ClassModel.DeclareClass(new RDFResource("ex:Cls"));
ontology.Data.DeclareIndividual(new RDFResource("ex:Idv"));
ontology.Data.DeclareNegativeIndividualType(ontology.Model, new RDFResource("ex:Idv"), new RDFResource("ex:Cls"));

• CheckIsNegativeIndividualOf: to check if the given individual is explicitly not of the given class (which was your original question)

bool isNotIdvOfCls = ontology.Data.CheckIsNegativeIndividualOf(ontology.Model, new RDFResource("ex:Idv"), new RDFResource("ex:Cls"))

If this is desired, I'll release a new version with the code in the next few days.

Regards,
Marco

[mdesalvo]

Hi constnick, given your individual I0 and your complement class C_C0, you can use:

bool isnotaC0 = myOntology.Data.CheckIsIndividualOf(myOntology.Model, I0, C_C0);

Regards,
Marco

[constnick]

It is checking for — is I0 explicitly belongs to C0 or not.
But I need to check, is I0 explicitly not belongs to C0.
For example in Turtle:

test:Man rdf:type owl:NamedIndividual ,
                   test:Human ,
                   [ rdf:type owl:Class ;
                     owl:complementOf test:Japanese
                   ] .

It means that Man is definitely known as not Japanese.
I need to check if there is such a declaration in the ontology.

[mdesalvo]

Hi,
although the library is answering your need, I understand that you have a slightly different semantic to prove. This kind of check must be done at SPARQL level, where you have full power of controlling the raw RDF triples of the knowledge in such way:

//Get down to RDF/SPARQL level
RDFGraph ontologyGraph = ontology.ToRDFGraph(true);

//Build an ASK query detecting individuals of any complement class, constraining the answer to the "test:Man" individual and the "test:Japanese" class to be targeted
//Maybe it is more verbose than needed, containing pattern which may be omitted, but I prefer to remain formal
RDFAskQuery isNotJapaneseQuery = new RDFAskQuery()
    .AddPatternGroup(new RDFPatternGroup()
        .AddPattern(new RDFPattern(new RDFVariable("?IDV"), RDFVocabulary.RDF.TYPE, RDFVocabulary.OWL.NAMED_INDIVIDUAL))
        .AddPattern(new RDFPattern(new RDFVariable("?IDV"), RDFVocabulary.RDF.TYPE, new RDFVariable("?IDVCLS")))
        .AddPattern(new RDFPattern(new RDFVariable("?IDVCLS"), RDFVocabulary.RDF.TYPE, RDFVocabulary.OWL.CLASS))
        .AddPattern(new RDFPattern(new RDFVariable("?IDVCLS"), RDFVocabulary.OWL.COMPLEMENT_OF, new RDFVariable("?NOTIDVCLS")))
        .AddFilter(new RDFSameTermFilter(new RDFVariable("?IDV"), new RDFResource("test:Man")))
        .AddFilter(new RDFSameTermFilter(new RDFVariable("?NOTIDVCLS"), new RDFResource("test:Japanese"))));
RDFAskQueryResult isNotJapaneseQueryResult = isNotJapaneseQuery.ApplyToGraph(ontologyGraph);

Console.Write("test:Man IS NOT test:Japanese? " + isNotJapaneseQueryResult.AskResult);

Let me know if this fits yor needs.
Regards,
Marco

[constnick]

Yes, certainly I can do what I want with SPARQL, but it would be convenient to have such methods in an ontology object or its component.
By the way:
There are two kinds of "spo" or "spl" assertions: positive and negative.
And also there are the same two kinds of "type" statements.
For example, in Manchester syntax it expressed just with simple "not":

Individual: test:Man
    Types: 
        test:Human,
        not (test:Japanese)

May be it would be consistent to have symmetrical tools to work with negative "type" statements, just like with negative spo/spl assertions?

[mdesalvo]

I didn't know the existence of the "negative type" statement. Maybe a OWL2 syntax which I didn't notice.

So, given an individual "ex:Idv" and a target class "ex:Cls" being the prohibited class, I'll figure how to add modeling support for this new construct:

ex:Idv rdf:type owl:NamedIndividual, 
                [ rdf:type owl:Class; owl:complementOf ex:Cls ] .

Regards,
Marco

[constnick]

Yes, I think that's what is needed, thanks!