I try to create an ontology in OWL, using Protégé 5.5.0. No I have a little trouble with inferred classes.
I have three classes: Agents and as subclasses Groups and Persons. Persons can be members of Groups. Now I want to create two inferred classes: a) "Members of groups" and b) "Not members of groups" (both as subclasses of Person)
I was successful with a), using the axiom equivalent class: Person and member_of some Group.
I created one individual, which is a person and member of a group and it was inferred to be member of the class "Members of groups".
Now I'm stuck with b). I tried several options for the equivalent class, for example:
Person and member_of max 0 Group
Or: Person and member_of exactly 0 Group
Or: Person and not member_of some Group
I created an individual which is a person but no groupmember. But the reasoner does not agree with me about the fact, that this individual should belong to the inferred class "Not members of groups".
What did I do wrong?
One way to achieve this in OWL, using your Person and not member_of some Group,
is to define a MemberOfGroup class that is defined as you suggested. Then define a NotMemberOfGroup class, that is disjoint with the MemberOfGroup class. If you now define your individual to be of type Person, as well as of type not member_of some Group, then your individual will be classified as belonging to the NotMemberOfGroup class.
The reason why you have to do this, is that OWL uses the open world assumption and hence it can only make inferences about things it knows for sure. I.e. saying that the individual is a Person, while making no statement regarding member_of provides the reasoner with zero explicit info to determine that the individual either belongs or does not to a group.
The other option is to use SHACL/SHEX/SPIN.
Related
I would like to create a new class PossibleNoduleFinding, grouping some individuals from another class Finding based on the individual name. Finding has a relationship hasFinding with each individual. Example maybe individual name nodule.
I'm trying:
Class: PossibleNodueFinding
Annotations: [in root-ontology]
rdfs:label "PossibleNoduleFinding"#en
SubClassOf: [in root-ontology]
SoftwareMapping
EquivalentTo: (Finding and hasFinding value nodule),
EquivalentTo: (Finding and hasFinding value density)
The Owl editor is not recognizing this syntax.
In protege a reflexive property is assigned to all individuals relgardless of domain and range and the class to which individuals belongs.
so what is the use of this restriction?
P.S: lets say there is three individuals:
NamedIndividual( :John )
NamedIndividual( :alex )
NamedIndividual( :BMW )
and an object proeprty:
ObjectProperty( :hasFriend )
ReflexiveObjectProperty(:hasFriend)
running pellet deduce that :
BMW hasFriend BMW
This inference is conceptually meaningless
Papers like The even more irresistible SROIQ and Foundations of Description Logics point out that reflexive and irreflexive properties are closely related to the exists r.Self concept. I.e. Narcissist can be defined as Narcissist \sqsubseteq loves.Self.
The SROIQ paper actually mentions that the main use cases for reflexive and irreflexive properties are limited and only make sense when used along with cardinality restrictions. I.e. if you define PopularPerson as someone with at least 2 friends, and hasFriend is reflexive, then by asserting an individual has 1 known friend will result in that individual being classified as a PopularPerson because the individual is already assumed to be its own friend.
Interestingly, the paper also mentions that reflexive(r) is equivalent to adding the GCI top \sqsubseteq exists r.Self to the TBox. Personally for me this is more intuitive and provides the control I think you seem to want to achieve. In particular this allows you to replace \top with whatever class of your choice. A similar equivalent exists for irreflexive properties.
In an OWL ontology, given a class Student; I want to define another class StudentsPair, which is any pair of students, in such a way that automatically computes all the possible dual combinations of Student, i.e.
Student(x) and Student (y) --> StudentsPair(x,y)
I want StudentsPair as a class, not a property, because it may have additional features (such as averagePerformanceForPair, etc.).
For example,
Premise
Student(John)
Student(Alex)
Student(Mary)
-----------------
Conclusion
StudentPair((John, Alex))
StudentPair((John, Mary))
StudentPair((Alex, Mary))
The reasoner has to somehow create these new individuals! Is this possible?
You need an additional role student, then you can use Description Logic Rules, something like that might work:
Student ≡ ∃student.Self
student ◦ U ◦ student ⊑ StudentPair
I'm using Protege 4.3 to create my first ontology, so I encountered a few problems
let's say we have the classes Person and Department and the properties employs and hasPart
So the problem is: I defined some classes like BigShopas Equivalent To employs min 6 Person and hasPart some Department but when I activate the reasoner I don't get the results I'm waiting for, I also tried SubClass Of with the same two restrictions but nothing happened
Any idea why?
In OWL, is it possible to query a class that does not have a property?
Suppose I have an object property R and I want to retrieve all classes that do not have the property R. Also, suppose that I have already closed all classes with closures.
I was trying this:
suppose the propety in question is hasProperty, my query was this
hasProperty only Nothing
But it doesn't work
What do you mean by "class that does not have a property"?
Semantically, a class is a set of individuals and a property is a set of pairs of individuals. Given these two sets, what do you mean by "class has / has not a property"?
Regarding your example query
hasProperty only Nothing
lets rewrite it a bit so that we can think about it in natural language. This gives better intuition about the meaning of this query.
First lets rename hasProperty to follows (or whatever English verb), then we have:
follows only Nothing
This is semantically equivalent to
follows only (not Thing)
which is semantically equivalent to
not (follows some Thing)
Now we have gotten rid of the only which is a confusing part of OWL and is better avoided. So now we can verbalize the query in English as:
those who do not follow anything
or
who does not follow anything?
or more formally
which individuals of the ontology are known
not to have a follow relationship to any other individual
E.g. if your ontology states that
John does not follow Mary.
John does not follow himself.
Every individual in the ontology is either John or is Mary.
then John will be the answer to the above query.
You can also get a named class as the query answer if the asked restriction holds for a (named) group of individuals. In any case, the following must hold: if you asserted that the answer (or a member of it) does have a follow-relation to some individual then it would render the ontology inconsistent.
OWL does not have any means for querying the data. SPARQL is used for that. So the SPARQL query to find all class definitions that do not have the property :R would be:
SELECT ?cls
WHERE {
?cls a owl:Class .
FILTER NOT EXISTS {?cls :R ?o}
}