Re: RDF and logic

On Wed, 4 Oct 2000, Tom Van Eetvelde wrote:

[snip]

> I know there has been some work around this (SILRI, DATALOG)
> but I wonder, do we need a separate logic layer on top of RDF
> or can we use RDF to define a logical layer? I tried the latter
> by introducing operator resources. This looks good in schema
> form, but writing software to support these constructs looks
> quite difficult. Simple example: if I introduce an AND node, do
> I AND adjacent nodes or the complete trees that originate from
> these adjacent nodes? So ad hoc rules need to support the
> schema on how to interpret the AND.

It depends on what you call ad hoc. From the RDFS point of view,
there is no way to express that (for example) an "AND" resource
can only contain 2 operands (assuming that you want to model
binary operators). More generally, it is often impossible to
capture the semantics of the new primitives you introduce at the
RDFS level. This, I think, is the job of the logic layer.

So, in that sense, a logic layer on top could be viewed as a
formalism whose primitives are directly mapped into RDF(S). Part
of the semantics of this formalism is then known at the RDF(S)
level, insofar as the primitives of the logic layer map directly
to core RDF(S) primitives, but the semantics of the newly
introduced primitives can only be understood outside the RDF
framework.

> Has anyone got some experience with this kind of modelling? Do
> I have to leave the RDF framework in order to introduce a logic
> layer or does anyone have a schema for logics that is easily
> supportable in software?

What you propose here is very similar to what is being done in
OIL. The approach there is to extend RDF Schema with additional
modeling primitives. We wrote a paper about this, which is
available online through (second time today I'm posting this
link, sorry about that):

    http://www.ontoknowledge.org/oil/papers/extending-rdfs.html


Regards,

Jeen
-- 
                               Vrije Universiteit, Faculty of Sciences
Jeen Broekstra              Division of Mathematics & Computer Science
jbroeks@cs.vu.nl                                    de Boelelaan 1081a
http://www.cs.vu.nl/~jbroeks        1081 HV Amsterdam, the Netherlands

Received on Wednesday, 4 October 2000 10:28:49 UTC