[rdfweb-dev] Some RDF stats

Dan Brickley danbri at w3.org
Thu Aug 7 09:50:21 UTC 2003


* Dan Brickley <danbri at w3.org> [2003-08-07 05:41-0400]
> * Martin L Poulter <M.L.Poulter at bristol.ac.uk> [2003-08-07 10:14+0100]
> > On Wed, 6 Aug 2003, Julian Bond wrote:
> > 
> > > There's rather a lot of tags in the wild that don't appear in the spec! 
> > > foaf:asshole, foaf:rudeThing and foaf:wakeUpSlacker look like they could 
> > > come in handy.
> > 
> > And foaf:gender!!!! It's a strangely abstract way of describing people
> > that doesn't have a mention of gender or sex (Okay, I realise that it
> > might be difficult to settle on a vocabulary because of many kinds of
> > borderline case, but it's not like a declaration is compulsory, nor that
> > we're going to mandate physical checks.)
> 
> Yes, we should do gender/sex. Question is which, and how to deal
> respectfully with corner cases. Probably calling people corner cases
> isn't a great start, oops. But yes, Something Should Be Done About
> This. Let's have a bit of discussion here then I'll add something and
> flag it as 'unstable' for a while to get further feedback.
> 
> The technical answer (Damian suggested this) is probably is to model the 
> value space as classes, state that 'Male' and 'Female' are mutually 
> disjoint, but not that they exhaustively span the possible values.
> 
> Problem is this doesn't look great in RDF syntax:
> <Person>
>  <foaf:name>Alice A</foaf:name>
>  <foaf:gender>female</foaf:gender>
> </Person>
> 
> vs
> 
> <Person>
>  <foaf:name>Alice</foaf:name>
> <foaf:gender rdf:resource="http://xmlns.com/foaf/0.1/female"/>
> </Person>
> 
> vs 
> 
> <Person>
>  <foaf:name>Alice</foaf:name>
>  <rdf:type rdf:resource="http://xmlns.com/foaf/0.1/Female"/>
> </Person>
>  
> which could be written (rather clinically)
> 
> <Female>
>  <foaf:name>Alice</foaf:name>
> </Female>
> 
> 
> 
> Looking at this, I am inclined towards the first option, and to say that
> 'male' and 'female' are expected values for that property, but not 
> restrict other possibilities there. We could also write inference rules
> such as:
>    
> 
>      ?x foaf:gender 'male'
>      implies (or is it entails? I can never remember)
>      ?x rdf:type wordnet:Male
> 

ps. if it looks like I changed my mind halfway thru this post, I sort of
did. I think the right answer is to use rdf classes, but the syntactic 
overhead of writing the instance data that way will increase the 'yuck'
factor, so using a simple text valued property (linked to a class behind 
the scenes for modeling tidyness) is probably the way to go. Comments?



More information about the foaf-dev mailing list