Hi,
I think Yves' proposal kind of hits our complexity-line. Do you remember
we drew something about the complexity that voiD is aiming to solve at
the right corner of the poster? That might give people some idea the
sort of the complexity we want to expose in voiD "at the starting point"
:)
I'm not sure - at first glance Yves' proposal seems both really clever
and really simple, and is good because:
- easier to publish and query than the reified style of riro
- only requires one extra property in voiD
However I suppose it does open up some questions:
- What advice do you give to publishers about selecting representative
samples from what may be a very heterogenuous dataset?
- From which triple stores / endpoints would it be possible to query for
datasets with a particular pattern in their examples? AIUI the idea is
that the example document is stored and/or retrieved as a named graph, and
I'm not sure how consistently this works across current stores and
endpoints since the SPARQL spec (as we discussed, coincidentally, at eswc)
gives some leeway as to how named graphs are implemented etc.
I think what could be useful at this point is some sample voiD
descriptions to try querying against etc - I hope to try this soonish
(today or tomorrow), and will ping the list for comments when I have
something.
Cheers,
Keith
I think this might be a bit too complex though. I also thought about
the following possibility: having something like:
:dataset void:example :example.
where :example is a small RDF document holding an example of what you
could find in this dataset. For example, I might have -
:example {
:both a mo:MusicArtist; foaf:made :rec.
:rec mo:available_as :playlist.
}
Then, it would be easy enough to SPARQL for "datasets that hold music
artists and related playlists".
Cheers!
y
--
--
