Re: RDF mismatch with frame-based systems?


Dan Connolly wrote:
> Up to the QL '98 workshop[1], I considered RDF[2] a hand-in-glove
> fit with frame-based reasoning systems. So I was suprised
> to see
>         "The schemas of RDF allows hte definition of
>         attributes, so called property types.
>         The property types are -- in contrast to
>         frame based languages like F-Logic-- general
>         in the sense that they do exist independently of
>         classes. Thus, it is not possible to give the same name to
>         different properties for several classes if they
>         have different value ranges or cardinalities."

This was discussed in the RDF Schema WG, whether property definitions
(i.e., "slots") are unique for any particular class or not. It wasn't my
personal preference, but the latter alternative prevailed because of (I
believe) simplicity of the underlying model.

Please note that there are different approaches to this in implemented
frame systems. For example, CRL (the Carnegie Representation Language)
is similar to RDF in that slot names are shared between different frame
classes ("schemata" in CRL terms). No mismatch there...

	- Ora

Ora Lassila, <>
Agent Technology, Nokia Research Center / Boston
phone: +1 (781) 238-4908

Received on Tuesday, 8 December 1998 14:19:22 UTC