(unknown charset) Re: another use case for RDF

On Thu, 1 Feb 2001, Al Gilman wrote:

> At 11:58 AM 2001-02-01 -0500, Dan Brickley wrote:
> >On Thu, 1 Feb 2001, Sean B. Palmer wrote:
> >
> >> > It's forwarded from teh RDF Interest Group. Since Sean is the

> >> > perpetrator, he has likely seen it already and can be clearer about
> >> > the goals, but the basic idea is dealing with spelling mistakes.
> >>
> >> Yes, but it's not valid RDF because DanBri's original letter wasn't in RDF.
> >> To get it to work, you'd have to convert an entire letter/page/+ to data,
> >> spell check it, and then convert it back - i.e. as part of a specialized
> >> spell-checker, not an RDF processor.
> >> Having said that, I do like the ideas from Charles/Lisa about using RDF for
> >> spell checking, a thesaurus, a dictionary, etc. DanBri's Wordnet stuff is a
> >> good indicator of what RDF can do, and it would be very easy indeed to
> >> modify it to become a thesaurus or a dictionary. Since Dan is now an
> >> official unlurker on the list, maybe he can chip in :-)
> >>
> >
> >
> >So WordNet is indeed a very interesting case. Unlike many Library
> >thesauri, the modelling has been done cleanly enough to project the
> >broader/narrower axis onto RDF's class hierarchy. This is because WordNet
> >makes distinctions such as:
> >
> >(excuse the ascii art rdf)
> >
> >    fido -- type --> Dog -- subClassOf --> Mammal
> >
> >rather than the looser
> >
> >    fido -- broaderTerm --> Dog --> broaderTerm --> Mammal
> >
> 
> Does RDF, itself, recognize [ -- type --> ] -- subClassOf --> [ -- subClassOf
> --> ] ?

Depends what one thinks of as RDF. At the starkest core, RDF is just a
bunch of edge-labelled graph data. However the RDF Model and Syntax spec
defines some basic utilities, such as rdf:type, and the RDF Schema spec
defines additions to these, fleshing out the class system a bit. So yes,
RDF (Model+Syntax and Schema) do have these notions 'built in'.

> 
> What is the value of distinguishing [ -- type --> ] from [ -- subClassOf
> -->] ?

It maps onto a reasonably intuitive model. In WordNet, they catalogue
the nouns hierarchies using a template of "A ___ is a kind of
___". Eg. "A _Dog_ is a kind of _Mammal_". With implied (er,
whadderyoucallit... transitivity), so 

fido <-type- Dog <- subclassof-- Mammal <-- subClassOf-- LivingThing

allows us to say that "Fido is a kind of Dog; all Dogs are mammals, all
Mammals are Living things". ... fido is a Living thing etc etc

So the rdf:type and rdf:subClassOf license different inferences. We
wouldn't say that "a fido is a kind of Dog" because fido (in my
example) is an individual, not a kind of thing.

BTW classic conflation of these two concepts is the word "isa" which can
be used in either sense...

So this stuff gets really fun when you stuff it inside JPEGs, PNGs
etc. You can say that an image depicts a (some-detailed-class) and have
implied class membership further up the hierachy. See
http://www.tasi.ac.uk/2000/09/rdfmeta/ for a sketch of this with PNG
images of Lions (interactive demo currently offline...)

Does this make any sense?

Dan

Received on Thursday, 1 February 2001 12:35:35 UTC