Re: Person and fictional Re: VideoGame proposal

Roughly.

It seems to me that currently items represented in schema.org, such as 
instances of CollegeOrUniversity, are actual things in the real world, so, for 
example, Hogwarts would not qualify.  This rule is actually strengthened by 
the wording at http://schema.org/Person
  Thing > Person
  A person (alive, dead, undead, or fictional).
which explicitly calls out that instances of Person can, exceptionally, be 
fictional.

Adding in a class like FictionalThing would change the above rule, making the 
exception be the rule (instead of proving the rule).


Adding expressive power to the schema.org formalism would permit getting 
around this conundrum, perhaps using higher-order functions and axioms as you 
note that Cyc does.

peter

PS:  I'm not sure that the "/" mechanism is a good one to point at.

PPS: I'm quite opposed to having Person include fictional people at all, as 
that makes Person be much too ill-defined.  Is a humming rock in a Disney 
cartoon a person?  What about a bouncing tree?


On 10/20/2014 11:38 AM, Simon Spero wrote:
> On Oct 20, 2014 8:16 AM, "Peter F. Patel-Schneider" <pfpschneider@gmail.com
> <mailto:pfpschneider@gmail.com>> wrote:
>
>  > The essence of these proposals is that there is some class or property that
> changes the meaning of something else. My worry is that producers and
> consumers will need to understand all such classes and properties before they
> can use schema.org <http://schema.org>.
>
> Would your issue be that if eg. Person is assumed to have some property of
> "not being an abstract entity", and that asserting "fictional" of such an
> individual would require defaults, since an individual that was Person and
> Fictional would be inconsistent.
>
> FOAF made the explicit decision not to make this part of the definition of
> Person. This does not necessarily help with the general case.
>
> One could augment the schema.org <http://schema.org> logic with something
> similar to Cyc's FictionalFN (and similar class generating functions).
>
> Since the precedent already exists for inferring meaning from the syntactic
> form of a class name ("/" as subclass/subproperty definition), one could use
> "fictional(Park)" as the class of an imaginary parks.
>
> The function could be declared as preserving subclass relationships between
> classes of fictional things ( fictional(Park) is a subclass of fictional(Place)).
>
> No subclass relationship between fictional(Park) and Park would be entailed.
>
> This approach would avoid having to manually add a fictional version of every
> possible class, allow fictional things to be ignored by applications that
> don't want to know about them, and allow special cases for applications that
> only care about a few kinds of fictional things.
> Nice and backwards compatible.
>
> Simon
>

Received on Tuesday, 21 October 2014 06:29:49 UTC