On 07/04/12 10:37, Ivan Herman wrote:
Markus et al,
what you are saying is true. However... the RDF Working Group that is currently in
operation will, hopefully, come up with a proposed syntax (probably based on TriG) and,
more importantly, some sort of a semantics for named graphs, hopefully in alignment with
SPARQL. I cannot say, of course, when this will be finalized and how it will align with
the timing of the Wikidata project. But it is worth knowing about it and, actually,
possibly to keep an eye on it and contact the WG if the (obviously important!) Wikidata
use case does not align with what the WG is doing.
Yes, we are aware of this activity and will be watching the outcome. One
rationale behind our abstract data model is that it makes it easy to
adopt future standards. Even if named graphs are not available yet at
the end of the project, it would be easy to write a new exporter later.
In this case, the old (triple) export and the new export would most
likely not have the same formal semantics, since they encode data in
different structures. But this is not a problem since we have a
technology neutral data model, that may well be faithfully represented
in different, mutually incompatible formats. In essence, this is our
extension strategy for supporting future W3C (and other) standards.
(And, of course, I am happy to do the go-between when and if the time comes:-)
Thanks. Once we have concrete proposals for an RDF/OWL export format, we
can also discuss how to improve it to make the best use of the available
language standards.
Best regards,
Markus
On Apr 6, 2012, at 17:31 , Markus Krötzsch wrote:
Martynas,
what you are proposing below is not W3C recommended RDF but an extension of triples to
quads. As far as I know, this extension is not compatible yet with existing standards such
as SPARQL and OWL. Named graphs work with SPARQL, but are mostly used in another way than
you suggest. Most RDF database tools would be *very* unhappy to get millions of named
graphs in combination with queries that use variables as graph names. The syntax you use
is not a W3C standard either.
This does not say that N-Quads aren't a good idea if one can get them to work with
the rest of the Semantic Web stack, but it really defeats your own arguments. We are
committed to supporting *existing* standards (as we have said many times already), but we
will not base our software design on a non-standard RDF-variant that works with neither
OWL nor SPARQL.
Markus
On 06/04/12 13:09, Martynas Jusevicius wrote:
Hey Denny,
I gave it a shot:
<http://dbpedia.org/resource/France>
<http://dbpedia.org/ontology/PopulatedPlace/populationDensity>
"116"^^<http://dbpedia.org/datatype/inhabitantsPerSquareKilometre>
<http://wikidata.org/graphs/France2012> .
<http://dbpedia.org/resource/France>
<http://dbpedia.org/ontology/populationDensity>
"116"^^<http://www.w3.org/2001/XMLSchema#double>
<http://wikidata.org/graphs/France2012> .
<http://wikidata.org/graphs/France2012>
<http://purl.org/dc/terms/date>
"2012"^^<http://www.w3.org/2001/XMLSchema#year>
<http://wikidata.org/graphs/France2012> .
<http://wikidata.org/graphs/France2012>
<http://purl.org/dc/terms/source> _:source
<http://wikidata.org/graphs/France2012> .
_:source<http://purl.org/dc/terms/published>
"2010"^^<http://www.w3.org/2001/XMLSchema#year>
<http://wikidata.org/graphs/France2012> .
_:source<http://purl.org/dc/terms/title> "Bilan demographique"@fr
<http://wikidata.org/graphs/France2012> .
The syntax is N-Quads. It does not use reification, but instead named
graphs for provenance. The necessary concepts were already present in
DBPedia.
As you might know, temporal provenance is not the strongest point of
RDF. However conventions and solutions are available, and I am sure
implementing them would require far less effort than creating a custom
data model from scratch, not to mention the benefits of potential
reuse.
There's quite some research done on RDF provenance, which is worth
looking into if provenance is really a key feature for Wikidata from
day one. I see it as something that should work transparently behind
the scenes, and therefore could be rolled-out later on.
You would get much better and more extensive advice than mine on
semantic-web(a)w3.org -- the only prerequisite is willingness to
cooperate.
RDF's strength is that it solves data integration problems by pivotal
conversion, reducing the number of model transformations from
quadratic to linear:
http://en.wikipedia.org/wiki/Data_conversion#Pivotal_conversion
A custom data model brings up questions which already have an answer
in the Semantic Web stack:
# can data from different Wikidata instances be merged or interlinked natively?
# is there a native query language? In case of SQL, how performant
will it be given many JOINs and the planned use of provenance?
# what and how many custom serialization formats and API mechanisms
will have to follow?
Stacking one custom solution on top of another can eventually result
in huge costs. I honestly think the energy of Wikidata could be
directed in a more productive way.
Martynas
graphity.org
2012/4/5 Denny Vrandečić<denny.vrandecic(a)wikimedia.de>de>:
Dear Martynas,
if you try to model the following statement in RDF
"The population density of France, as of an 2012 estimate, is 116 per square
kilometer, according to the "Bilan demographique 2010"."
you might notice that RDF requires a reification of the statement. The data
model that you have seen provides us with an abstract and concise way to
talk about these reifications (i.e. via the statement model, just as in
RDF).
We still have not finished the document describing how to map our data model
to OWL/RDF, but we have thought about this the whole time while discussing
the data model.
But if you find a simpler, and more RDFish way to express the above
statement, please feel free to enlighten me. I would be indeed very
interested.
Cheers,
Denny
2012/4/5 Martynas Jusevicius<martynas(a)graphity.org>
>
> it doesn't look like reuse of existing concepts and standards is a
> priority for this project.
> One cannot build a Semantic Web application by ignoring its main
> building block, which is the RDF data model.
--
Project director Wikidata
Wikimedia Deutschland e.V. | Eisenacher Straße 2 | 10777 Berlin
Tel. +49-30-219 158 26-0 |
http://wikimedia.de
Wikimedia Deutschland - Gesellschaft zur Förderung Freien Wissens e.V.
Eingetragen im Vereinsregister des Amtsgerichts Berlin-Charlottenburg unter
der Nummer 23855 B. Als gemeinnützig anerkannt durch das Finanzamt für
Körperschaften I Berlin, Steuernummer 27/681/51985.
_______________________________________________
Wikidata-l mailing list
Wikidata-l(a)lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l
_______________________________________________
Wikidata-l mailing list
Wikidata-l(a)lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l
_______________________________________________
Wikidata-l mailing list
Wikidata-l(a)lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l
----
Ivan Herman, W3C Semantic Web Activity Lead
Home:
http://www.w3.org/People/Ivan/
mobile: +31-641044153
FOAF:
http://www.ivan-herman.net/foaf.rdf
_______________________________________________
Wikidata-l mailing list
Wikidata-l(a)lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l