Act of Communication Subclasses

[swartik]

There seem to be two categories of subclasses for Act of Communication. The first category has to do with different communication intents: commissive, deceptive, declarative, directive, expressive, personal, and representative. The second concerns how the communication occurs: by media or by mass media (or both, perhaps).

Should all these classes be siblings? Or should the CCO class hierarchy reflect the categories? I expect almost every individual that is a member of class Act of Communication by Media is also a member of some Act of Communication subclass in the first category. For example, mailing a birthday card is both a Mailing and an Act of Congratulating.

Also, if @alanruttenberg's point about Act of Communication by Media subclasses in issue #705 is accepted, most of that class's subclasses will move up a level. That would make for a large number of Act of Communication subclasses.

[alanruttenberg]

There's no ontological principle that cares about "too many siblings". Subclass means something exact: If a is subclass of b, then all instances of a are instances of b. Alternatively, if you have a class c, then every instance of c must conform to all the definitions of the transitive superclasses up to Entity. The OWL 2 direct semantics governs this. That there is this clear definition of what the child/parent relationship means is one of the things that distinguishes some arbitrary taxonomy or thesaurus from an ontology.

The only reason to interpose a new class to reorganize a list of siblings is if the the new class captures a generalization: something that is true of all the now subclasses. It's not like adding a new folder to help organize your files into neater piles.

If it is the case that in some cases where instances of subclass A of Act of Communication by Media is also a subclass of some subclass B of Act of Communication, and A subclassOf B is not reflected in the subclass hierarchy then it is also a mistake. There are many such mistakes in CCO, for example #232.

The way I put this sometimes is that you don't have any choice about what is subclass of what. You only have a choice of what terms to include in your ontology and how they are defined. Then, as a matter of necessity, what is subclass of what follows from the definitions and what it means to be a subclass.

Now if it happens that there is a missing subclass relation - a case where definitions imply subclass but the ontology does not say they are subclasses - then the fix, according to good ontology practice is not to assert both superclasses, but rather assert one and give the other a definition using equivalentClasses such that the reasoner will infer the second subclass relation. In many cases the presence of a situation of like this means something is not factored properly. A refactoring takes some aspect of the definition and builds a hierarchy of those differentia, and then reuses or introduces a new relation so that a logical term definition can be made composing the two classes - the original and one of the members of the new hierarchy. Here is a representative article on the technique, called Rector Normalization.