How would you model that a potential relationship exists between two classes, but doesn't always?

I want to model that something may be, but is not guaranteed, to be true. For example, a house may have a garage, but not necessarily. And a garage may be part of a house, but not necessarily.

If I add as an axiom "has_part some 'garage'" to my 'house' class, I'm claiming that all houses have garages, which isn't always true. Similarly, I can't add "partOf" to my garage class for the same reaosn.

What would be the ideal way to model this relationship?

I've tinkered with declaring that the class 'house' "has_part some 'garage' min 0" in order to use cardinality to set the minimum bound to zero, but I'm not sure that's correct.

Solution

In that case you you can think of the class House consisting of two different classes: (1) the class of houses that do not have garages and (2) the class of houses that do have garages.

You could do this in Manchester syntax as follows:

 Class: ClassWithoutGarage
    SubClassOf: hasGarage max 0 Thing //has no garage

 Class: ClassWithGarage
    SubClassOf: hasGarage min 1 Thing
    DisjointWith: ClassWithoutGarage //has at least 1 garage

 ObjectProperty: hasGarage
    Domain: ClassWithGarage
    Range: Garage

 Class: House 
   EquivalentTo: ClassWithGarage or ClasWithoutGarage

In this case we have given explicit names to these 2 possible classes. Often in ontologies, the existence of these classes can be implied without explicitly naming them:

Class: House
   EquivalentTo: hasGarage max 0 Thing 
                 or  
                 hasGarage min 1 Thing

which can be stated more concisely with

Class: House
    EquivalentTo: hasGarage only Garage

hasGarage only Garage defines the class of individuals consisting of those individuals x such that when individual y is linked via hasGarage to individual x, then y is of type Garage, or those individuals x that has no link via hasGarage.