OWL2 and FOAF

Hey Bijan, (cc: others who might be interested,)

Good to see you last week at TPAC. Hope you enjoyed the rest of the meeting.

I'd like to pick up the discussion about OWL2 and FOAF.

What tools can I use to help migrate http://xmlns.com/foaf/0.1/index.rdf 
into a more OWL-friendly form? Is there a tool that tracks the current 
state of the OWL2 spec?


Off the top of my head, issues I'm concerned with are:

1.
migrating from IFPs on literal valued properties (mainly 
foaf:mbox_sha1sum) to use easy keys
- do these work when applied to bnodes?
- is there any way to say that we want identity reasoning to ignore 
xml:lang values on these?

2.
foaf:Group descriptions
Currently we have foaf:membershipClass,
http://xmlns.com/foaf/0.1/#term_membershipClass
...this relates a Group to the OWL class that picks out the subclass of 
Person who are a foaf:member of it.

This seems to be largely unused in its current form. I have been tempted 
to add in another mechanism for populating group descriptions 
automatically via SPARQL queries. For example,

CONSTRUCT {
?x foaf:member http://example.com/w3cstaff.rdf#itself
}
FROM NAMED <http://www.w3.org/People.rdf>
WHERE {
?x foaf:workplaceHomepage <http://www.w3.org/>
}

ie. written down queries that bake in some notion of source/provenance 
of the claims: the "W3C Staff" FOAF group is the group of people that 
some specified w3.org page says works for W3C. I understand this is 
tricky in OWL as-is, but can at least be encoded in SPARQL in a way that 
should run directly against crawler datasets. Advise on this would be 
very welcome.

3.
Inter-spec dependencies. Wherever the FOAF RDFS/OWL uses dublin core 
stuff, we get dragged into OWL Full, unless we go declare DC to be using 
annotation properties. This arose partly when using dc: internally to 
document the FOAF vocab. But also now we have a more ontologically 
inclined version of dc, http://purl.org/dc/terms/, it is likely to crop 
up if we express relationships between eg. our notion of Agent and 
dc:Agent. Including such links in the OWL seems a healthy thing to do, 
but it would be a shame if this meant we couldn't run FOAF data through 
OWL2 tools properly.

It's a while since I ran FOAF against an OWL validator, so there are 
probably other issues. I did some years ago try to minimise the things 
that kept us in Full, but clearly there's more to do.

I know FOAF's documentation, for better or worse, gets widely copied 
when people bash out little RDF vocabs. So I'll love to get it using OWL 
2 as soon as possible...

Thanks for any pointers!

cheers,

Dan

}

Received on Monday, 27 October 2008 14:48:37 UTC