Re: What do the ontologists want

From: "pat hayes" <phayes@ai.uwf.edu>

> Modelling propositions in graphical form is on old game, arguably
> pre-dating modern logic. Peirce's original calculus (from the late
> 1800s) is graphical, and many other graphical notations have been
> devised and are in widespread use. To the best of my knowledge, none
> of them use reification, however. I fail to see how reification
> arises naturally in graphical modelling; again, this seems to me to
> be based on a confusion between thinking of a graph arc as indicating
> reference, and thinking of it as indicating syntactic structure.

Hmmmm .... i wonder if we can just take KIF off the shelf ?

A ~labeled~ graph arc is a syntactic structure.  It seems to me that a
inference engine can be constructed to behave according to whatever meaning
is designated for its ~label~ in the graph.  So that  If we have a triple
{Seth isReferedToBy "Seth"} and also have in the graph {isReferedToBy
subProperty Reference; domain Person; range Literal} there would be a way to
attach a KIF definition of Reference to {Reference} in our graph as follows
{Reference KIFdefinition (.....put actual KIF definition of Reference here
...)}  such that our inference engine would behave properly according to
it's axioms and rules for reference when it encounters the original triple
above.

Would that work?   If not why not?  If so, does it not show at least one way
to connect the RDF data model (which I take to be just labeled directed
arcs) with Model Theoretic Semantics ?

Incidentally I am aware that this is not an example of reification, but
hopefully it cuts closer to this basic misunderstanding between us.  I would
draw a mentograph, but you don't respond to them - so I have asked my
question in a string of words.

Seth

PS:  For 'behave' you can substitute 'generate true statements' if that is
more appropriate.

Received on Wednesday, 16 May 2001 17:05:00 UTC