Re: Meaning of an RDF document: issue rdfms-assertion

Your (3) seems to override (2); or rather, you allow
that (2), "predicates determine meaning" to devolve meaning of cetain
statements to the meaning of the object of those statements. If that's the
case, are there cases where (2) will devolve meaning-setting rules in
other ways? (eg. consider an inverseProperty of rdf:type). In which case,
(2) doesn't seem such a forceful observation. Its the predicate that sets
the meaning, unless the predicate says that something else sets the


On Fri, 16 Nov 2001, Tim Berners-Lee wrote:

> It is important that the specification (at W3C) of a language
> ground the meaning of a document written in that language
> in terms of a technical specification.
> In the case of RDF, there is a key step, well 3 key steps
> in the meaning of an RDF document which pass through the
> RDF and RDFS specs.  Most of the meaning is then defined
> by the specfications of various terms in ontologies. But
> it is important that the three steps below are not omitted
> from the RDF specification.
> 1.  The meaning of an RDF document is the sum
>  (in english, strictly conjunction) of the independent meanings
>  of the statements of which the RDF document is comprised.
>  (Here we are talking about an RDF document which contains
>   directly.  RDF embedded in XML or anything else
>   only has meaning in as much as it
>   is given meaning by the specification of the language which
>   envelops it.)
> 2. The meaning of an RDF statement is defined by the
>   preciate used, and so is specified by the specification
>   of the Property that is used as predicate.
> 3.  When the predicate of  statement is rdf:type, then the
>   meaning of the statement is defined by the class used
>   as object, and so is specified by the specification of the
>   class.
> They are obvious if you think about it, and represent how the
> world actually works.  It is important to emphasize that in RDF
> one cannot, for example, define a special object which modified the effect
> of the predicate.  (By contrast, you can define in XML a new attribute
> which negates the effect of a tag... and then define a new tag which
> works un-negated even with the attribute and so on.  There are
> no axioms.)
> (Note that there is no meaning here associated with RDF embedded in random
> XML.  I think this needs to be addressed. You actually need a schema or
> document annotation language  to be able to get over this where you want to.
> So if you have a longer term issues list then that should be on it.)
> Tim Berners-Lee
> PS: This, with a couple of other things,  was discussed on #rdfig today.

Received on Friday, 16 November 2001 16:58:54 UTC