[rdfweb-dev] [ANN] Rosco, a non-judgemental RDF document checker

Ian Davis iand at internetalchemy.org
Fri Aug 15 13:10:23 UTC 2003


On Friday, 15 August 2003 at 13:04, Libby Miller wrote:
> It's a generic tool, although the examples are from FOAF.

> I hope it's useful - feedback welcome.
This is what I was doing with the Semantic Planet crawler a while
back. However, only about 1% of all FOAF out there was explicit enough
to be considered 'valid'. Once I realised that the domain/range
properties were inference rules then the notion of validity became
less important.

For those who haven't come across this it lets you write even simpler
RDF:

<rdf:Description rdf:about"bah">
  <foaf:knows rdf:resource="humbug" />
</rdf:Description>

This gives the following triple:

<bah> <foaf:knows> <humbug>.

But an RDF Schema aware parser should also add the following to the
model:

<bah> <rdf:type> <foaf:Person> .
<humbug> <rdf:type> <foaf:Person> .

So we've learnt something about humbug without having to fetch it.

To get the same triples from a non-RDF Schema aware parser the author
would have to write:

<foaf:Person rdf:about"bah">
  <foaf:knows>
    <foaf:Person rdf:about="humbug" />
  </foaf:knows>
<foaf:Person>

I could also add my own private schemas that add different inferences.
Perhaps in my world the fact that one resource 'knows' another implies
that both resources are CognitiveEntities rather than the more
restrictive foaf:Person. I don't have to apply the foaf schema if I
don't want to for my private processing requirements.

So, anyway, I'm in the process adding a lint facility to the SemPlan
crawler to pick out best practice and make suggestions on how to
produce better RDF, using Dan's 'polite and informative' metaphor from
http://rdfweb.org/topic/FoafNews



-- Ian <iand at internetalchemy.org>
"If we knew what it was we were doing, it would not be called research, would it?"




More information about the foaf-dev mailing list