Re: Inference rules [was: extracting...]

Dan Connolly wrote:
> wrote:
> > I see you making variables and rulesteps URI global.
> I'm not sure I understand what you mean by that.

Well, I saw you using #who, #t, #group and also #if1, #if2
#then which are relative URIs I suppose, that's all.

> Ah... thanks. I have glanced at that syntax a few
> times without understanding. Now I get it. Er...
> no, I don't: that's not legal RDF syntax, is it?
> the junction propertyNode can only have one child
> element, right?

I'm also not sure I understand what you mean.
The junction is a class name and not a property.
Although I see that we get things like literal("var:t") ... hmm ...

> > One can use the var namespace to have variable predicates
> > which can be useful for higher order logic such as in
> >
> > I guess this must be possible in shoe-swell as well,
> > just wonder how ?-)
> it's pretty straightfoward; see the "transitive"
> example in :
> r (x, z)
>      if
>           isa(r, transitive-relations)
>           r(x , y )
>           r(y, z)

Is it also possible to have things like
r(x, y)
           x(y, z)
           p(z, u)
where x stands for an unknown arc?

> Recognizing variables by their spelling is one
> technique, but it rubs me the wrong way... URIs are
> supposed to be opaque (not to mention the illicit
> use of an unregistered var: URI scheme). So I model the
> fact that a URI is used as a variable in a formula explicitly
> using the vars thing (which reduces to a forAll
> construct... see the "Relationship to FOPC"
> section of .../inference).

I see your point, thank you!

> Hmm... I wonder if I can express your model for rules
> in terms of the FOPC schema I'm working with...
> I think so.

I'm really looking forward to that!
Jos De Roo

Received on Monday, 31 July 2000 12:18:14 UTC