> It would be awkward in RDF, as well, since a membership qualified by a
> number is inherently a three-part relationship, which you do not
> directly have in RDF. You could create a node of type "membership" -
>
> membership
> member dog
> degree .9
>
> Then you could create a "fuzzy set" resource - maybe "pet" in this
> example - that is related to the membership node.
An alternative is to turn it around and apply a "certainty factor" to the
statement using reification:
{dog memberOf pets} certainty .9
But like you say, it's still pretty awkward.
> John Sowa has written that it takes three steps to moel the world, not two
>
> 1) A lattice of theorems (i.e., a purely mathematical structure)
> 2) A model, that stands between the world and the lattice.
> 3) The world itself.
>
> In this scheme, the model maps to the lattice of theories in a binary,
> true/false manner, while the model maps to the world in a fuzzy way.
> Thus, OWL classes would map between theorems and the model. There would
> be no need to try to force the model to map to the world in a crisp,
> true/false way.
>
> I don't know if this has been helpful, but at least it is interesting.
Yep.
Cheers,
Danny.
