- From: Pierre-Antoine <pierre-antoine.champin@liris.cnrs.fr>
- Date: Mon, 23 Nov 2009 13:28:59 +0100
- To: Tobias Bürger <tobias.buerger@sti2.at>
- CC: "Evain, Jean-Pierre" <evain@ebu.ch>, "public-media-annotation@w3.org" <public-media-annotation@w3.org>
- Message-ID: <4B0A800B.1020104@liris.cnrs.fr>
Tobias, conceptually, I am considering subproperties in the sense of RDF(S), i.e. composer (e.g. in the sense of ID3) is a subproperty of ma:contributor. syntactically, I suggest indeed to attach this information to the return value, so that users are free to use it or not. So md.getValues("contributor") would return like [ { "value": "John Doe", "subproperty": "id3:composer" }, { "value": "Jane Doe", "subproperty": "id3:lyricist" }; ] In a sense, this amounts to use the following SPARQL query (on some conceptual RDF graph, which may not be the way it is actually implemented, don't get me wrong!) : SELECT ?value, ?subprop WHERE { the:media ?subprop ?value . ?subprop rdfs:subPropertyOf ma:contributor . } I do not envision that the argument of our unique method could/should accept a subproperty, e.g. md.getValues("id3:composer") because I think this would be make immplementation quite harder (maybe too RDFish). I would prefer something like md.getValuesFilterSubproperty("composer", "id3:composer") But I'm ready to discuss the alternatives. pa Le 23/11/2009 09:40, Tobias Bürger a écrit : > Hi PA, all > > apologies for stepping in late in the discussion. > > If I understand your proposal correctly then you suggest not to define > subproperties in the sense of RDF(S) which allow subclass-relationships > among properties, but you suggest to define attributes for something > which is in the range of a property such as a contributor. > Do I understand this correctly? > If yes, then we are not discussing subproperties but structured return > values here (as we did in the early days of the working group). > > Thanks in advance for clarification. > > Best regards, > > Tobias > > Pierre-Antoine wrote: >> Le 20/11/2009 10:16, Evain, Jean-Pierre a écrit : >> >>> PA, >>> >>> do you mean a property/sub-property like title / (title) type? >>> or contributor / role? without specifying what the type or role >>> is but allow mapping to what is available from other descriptions. >>> >> >> Basically, yes, this is what I mean. >> More precisely, I suggest that, e.g. >> >> md.get("contributor") >> >> would return a set of values. Those values would basically be text, >> but would have an optional attribute (call it "role" or >> "subproperty"...) indicating more precisely the kind of contributor >> represented by the text. >> >> This optional attribute would represent additional semantics (w.r.t. the >> general semantics of ma:contributor), provided by the underlying format. >> At first, we can leave this field completely unspecified and let >> implementators do whatever they see fit to fill it. Later on, we could >> identify a set of standard values for these fields, to reflect notions >> that are considered relevant enough, and present in one or several >> underlying format. >> >> Again, try out to my implementation [1] (quite outdated regarding our >> drafts, but this is not the point here) for an example of this idea. For >> the moment, my implementation only provide the additional information if >> you explicitly ask for "structured" value. The sub-property is carried >> by the "property" field (quite ill-name, I agree ;)... >> >> My point is : we should decide now how to make this information >> available in the interface (the "structured" flag is not necessarily the >> good way to do it). This is a little extra work, granted, but it paves >> the way for extensibility (even if we chose not to standardize this >> extensibility -- de facto standard could as well emerge from this >> feature). >> >> pa >> >> [1] http://champin.net/wsgi/mawg/ >> >> >
Received on Monday, 23 November 2009 12:29:33 UTC