LANG: feedback on owl proposal


quick restatement of the points I made yesterday at the telecon:

1) 100% in favour of the approach (owl-frame vs full-owl)

2) 'full cardinality' should be in owl-frame

3) full set of property specification properties shoud be in 
owl-frame (symmetric, transitive, inverseOf, uniquelyDefining, 
unambigous, etc..)

4) drop 'primitive class & defined class terminology. I have not been 
able to come up with better alternatives. , so maybe we could use a 
different approach and qualify the attribute specs, rather than 
classes.  For instance

PrimitiveClass (male, supers(animal))


Class Male (attributeSpec supers(animal))

DefinedClass (Man, supers(Person, Male))


Class Man (completelydefiningAttributeSpec supers(Person, Male))

A possible extension of this approach could be to allow users to mix 
definitional properties with other properties in the same definition 
- i.e., attributeSpec and completelydefiningAttributeSpec could also 
be used in the same definition.  For instance:

Class Man (completelydefiningAttributeSpec supers(Person, Male),
                   (attributeSpec slot (hasAge, range=PositiveInteger)))

The above class spec to be interpreted as stating

Man (x) <-> (Person (x) and Male (x))
(Man (x) and hasAge (x, y)) -> PositiveInteger(y)

Another example:

Class DogOwner (completelydefiningAttributeSpec supers(Person)
slot (hasPet, range=Dog))

which means

DogOwner(x) <-> (Person (x) and hasPet (x, y) and Dog (y))

A possible advantage of this approach is that we can use DL-style 
specifications without giving up frame-style notation, thus avoiding 
the daml+oil style of specifying slots by saying that something is a 
subclass of another class with the slots in question.


plus an additional organizational point.  Would it be possible to 
include the other members of teh subgroup in offline discussions? 
Otherwise not sure what is the point of being in one group rather 
than another one.

Received on Friday, 22 March 2002 14:33:16 UTC