From: Stuart A. Y. <sy...@gm...> - 2008-12-04 06:29:13
|
On Thu, Dec 4, 2008 at 1:53 AM, Richard Light <ri...@li...> wrote: > In message <493...@ou...>, Sebastian Rahtz > <seb...@ou...> writes > >>some of you may be interested in some work here, >>pointed at by >>http://oxforderewhon.wordpress.com/2008/11/28/rdf-and-the-time-dimension >>-part-1/, >>worrying about temporal aspects of RDF. Its not a million >>miles away from the TEI. > > Interesting discussion. Your hemispheres/August example seems to me to > be a simple case of a conditional fact: IF you are in the Southern > Hemisphere, THEN August is a winter month. A quick Google for "rdf > conditional" turns up a W3C position paper from 2002: > > http://www.w3.org/2002/03/semweb/conditional > > suggesting (a) that this is seen as an actual issue by RDF full-timers > and (b) that it hasn't been resolved. > > This is indeed disappointing for me, when the sort of museum/historical > information I am interested in will be littered with statements which > are only true in a given context. You mention date ranges: many > assertions will only hold within a given range of dates (and that's > before you start to worry about uncertainty of dating ...). In > addition, assertions may be one person's opinion: the (varying) > attribution of art works is a good example. You need to be able to > express this sort of context clearly and (ideally) elegantly. > > It suggests to me that Stuart maybe shouldn't be in too much of a hurry > to move away from Topic Maps, which have the notion of Scope built in. > This seems much more powerful and elegant to me than named graphs (and > is actually part of the relevant standard). My understanding is that we only use scope for the language of labels (I've not looked at the i18n of RDF yet), so he inability inconsistent data isn't much of an issue or us, as I understand it. I'm sure my colleagues will jump on in if I'm wrong about this. > If you look at the sort of Linked Data which is coming out of the > DBpedia project, you see very simple metadata-like assertions about each > entity of interest, i.e. simple RDF triples with no blank nodes: > > <person X URI> <was-born-in> <birthplace Y URI> . > <person X URI> <isa> <person URI> . > <person X URI> <was-born-after> "earliest birth date" . > <person X URI> <was-born-before> "latest birth date" . > > This is the sort of stuff that a SPARQL end-point can really get its > teeth into. > > My problem is that the single DBpedia-like property <was-born-in> maps > to the chain: > > <P98B.was_born> <E67.Birth> <P7F.took_place_at> > > of CRM classes and properties. <was-born-after> maps to: > > <P98B.was_born> <E67.Birth> <P4F.has_time-span> <E52.Time-Span> > <P79F.at_some_time_within> <E61.Time_Primitive> <claros:notBefore> > > So in what way is the CRM useful for deriving these simpler "short-cut" > properties that the Linked Data initiative needs? The OWL language is ideally placed to express these "short cuts." OWL enables the introduction of purely synthetic vocabularies, i.e. vocabularies that are syntactic sugar for existing ones. http://en.wikipedia.org/wiki/Web_Ontology_Language Most topics present in the NZETC topic maps need not make it to the "short-cut"-ified version. We're aware, for example that dbpedia doesn't make a distinction between different editions, folios and printings of "Richard II" which CIDOC CRM does. Topics for these would be merged. OTOH, dbpedia does make a distinction between "Richard II" (the work) and "Richard II" (the king of England). cheers stuart |