Lists Home |
Date Index |
> Wrong model - not() is a monadic relation, not binary. RDF can have no
> monadic relations, even with [name your favorite] layered on top.
A is equivalent to not B
is a usable (binary) form of negation surely?
At risk of being mocked by logicians (the horror!) I would have thought that
it was possible to 'downcast' binary relationships using the universal,
A isFalse B
with B being false forall A
(Are datatypes monadic?)
> In RDF it is painful at best to refer to some particular GC-like
> context (an
> arbitrary subgraph), because subgraphs are not identifiable. You have to
> invent something equivalent to rdf:Statement, but more general.
Agreed, this bit is pretty horrible. I think there may still be hope
though - Graham Klyne's done some interesting stuff with contexts, and Alex
Barnell's RDF Objects  look promising.
> Anyway, the degree of extension to RDF needed would be significant. You
> need negation and quantification, at least.
The extension needed might be significant, but that doesn't mean it has to
be difficult, and even if it is difficult, it doesn't mean it wouldn't be
worthwhile. Then of course there's the issue of interfacing with the rest of
I believe TimBL's cwm has at least some support for negation and
quantification; I've been playing with RDF in Prolog a bit  (with some
quantification, negation and closed world assumption) and from a practical
point of view it's been very straightforward (well, apart from a few loops
because I haven't implemented tabling yet).
Anyhow, thanks for the interesting points. Believe it or not I remain
optimistic that RDF can do a lot of the stuff CGs can do without too much
extra work. I want a little relaxation this coming month, maybe I'll have a
crack at Analogical Reasoning in RDF ;-)
PS. Not read it yet, but this is another paper looking at RDF-CG mapping :