On Wed, February 12, 2014 07:14, Matthew West wrote:
Take a Person for example, with subclasses Boy and Man. [MW>] The main
problem with this is that Boy and Man are not subtypes of person. ...
To most people, and dictionaries, Boy and Man are subtypes of Person.
[MW>] If that were true it would only mean that most people were wrong.
No. Just that most people are not 4D fundamentalists.
OED Has:
[for "boy"]
A male child or youth. Also: a son, irrespective of age (chiefly as
referred
to by members of the immediate family).
Thus a boy is a type of child
And for child we get:
A young person of either sex, usu. one below the age of puberty; a boy or
girl.
Thus a child is a type of person, specifically one to which the modifier
"young"
applies.
In neither case does it say that a child is a subtype of person.
The OED does not use the word "subtype" or even "type" in its definitions.
The OED does not even use sentences in its definitions, so to a purist,
i guess, the definitions of the OED mean nothing since they are not
assertions.
However, a dictionary gives <modifier> <noun> as a definition for a term,
as long as the modifier isn't negatory ("former", "artificial", ...) the
meaning
of the definition is "an instance of <word> is an instance of <noun>
specifically
an instance of <modifier> <noun>". Thus the class of all <word> is a subtype
of the class of all <noun>.
Second, should a KB contain both a Boy & Man resource about a given
individual, owl:sameAs would be used to indicate their equivalence
otherwise, yes, they would be a different person, as they should be.
[MW>] Well that is precisely the problem with boy and man being subtypes
of
person. Now let us suppose that boy A has a start date of 1990-05-03 and
end
date of 2005-05-03 and the man B has a start date of 2005-05-03, and an
end
date of 2014-01-15. We now wish to say that boy A and man B are the same
person, and decide that OWL:sameas will do. What now is the start date for
A=B?
ManB, BoyA, and the person they are temporary slices of all have the same
birth date.
If the man has a start date of 2005-05-03, and an end date of 2014-01-15,
he has not aged enough by his end date to be a man by that point.
If you are using start date to mean the date when an instance became an
instance of that class (e.g., Man) then it isn't a property of the
individual,
but a property of the individual with respect to the class, and the
individual
would have different start dates with respect to different classes.
So
* (startDate BoyA Boy 1990-05-03)
* (startDate ManB Boy 1990-05-03)
* (startDate BoyA Man 2005-05-03)
* (startDate ManB Man 2005-05-03)
If we define "Boy" as starting at whenever "Person" starts:
* (startDate BoyA Person 1990-05-03)
* (startDate ManB Person 1990-05-03)
Third, StateofPerson is a wholly artificial term, lacking both practical
merit and semantic credibility.
[MW>] State of person is a noun phrase which has quite a precise meaning.
Which meaning can be quite different in different ontologies. In a medical
ontology "State of person" might be a set of measurements, diagnoses, and
other data, which can vary from second to second and minute to minute. In
such an ontology Man and Boy would not be subclasses of "State of person"
but merely features of such a state.
Fourth, this is a fine example of ontologists' implicit saintliness
modelling 'concepts' not 'language'.
[MW>] As others have said, I should hope so too. Ontology is about the
kinds
of things there are and the rules that govern them, not about the language
we might use to describe them.
Agreed.
-- doug