Here's my take.
RDF standards, in themselves, don't address all of the issues needed in a data wiki. I've been thinking about the math for data wikis and it seems to me you could have a bipartite system where you have "the fact" and then "the operational metadata about the fact" and these are conceptually two different things. Then you can query against the operational metadata to project out RDF or similar facts.
Had the Wikidata people had the right idea at the time and had good luck, they might have been able to build such a system. As it is they came up with a plan they were able to execute and they did it well.
The real trouble with an RDF export from Wikidata is that if you do a complete export, you're going to get something tangled up that you can't query easily with SPARQL.
There are really two answers to this, one of them the essentially forward-chaining approach of "create a canonical data model" (it doesn't need to be the "one ring to rule them all" for the whole web, just the one you need for your job), then you can project out an RDF graph where you've expressed your opinions about all the opinions in Wikidata.
There's a backwards-chaining kind of strategy where you, effectively, try running the query multiple times with different strategies and then do data clean-up post query. That's an interesting topic too, one that again demands something beyond ordinary RDF. Since RDF and SPARQL are so well specified it also possible to extend them to do new things, such as "taint" facts with their original and propagate it to the output.
I think also people too are realizing ISO Common Logic is a superset of RDF and it is really about time that support for arity > 2 predicates comes around. Note that arity>2 is already exists in W3C specifications, in that the fundamental object in SPARQL is a "SPARQL result set" which is an arbitrary-length tuple of nodes. It is clear what should happen if you write a triple pattern like
{ ?s ?p ?o1 ?o2 ?o3 . }
This also gives a more direct mapping from SQL to SPARQL, one that would be comfortable if there was some syntactic sugar to specify fields by names.
Yes, you can fake it by writing triple patterns, but in practice people struggle to even get simple SQL-like queries to work right, and can't do the very simple things people did with production rules systems back in the 1970s.
OWL was designed on the basis of math, not on the basis of "what are the requirements for large scale data integration". Thus it lacks very basic facilities, such as numeric conversions between, say, heights, in different units.
On Thu, Oct 30, 2014 at 9:08 AM, Kingsley Idehen kidehen@openlinksw.com wrote:
On 10/29/14 5:59 PM, Lydia Pintscher wrote:
Hey Phillip :)
On Wed, Oct 29, 2014 at 7:41 PM, Phillip Rhodesmotley.crue.fan@gmail.com motley.crue.fan@gmail.com wrote:
FWIW, put me in the camp of "people who want to see wikidata available> via RDF" as well. I won't argue that RDF needs to be the **native**> format for Wikidata, but I think it would be a crying shame for such a> large knowledgebase to be cut off from seamless integration with the> rest of the LinkedData world.>> That said, I don't really care if RDF/SPARQL support come later and> are treated as an "add on", but I do think Wikidata should at least> have that as a goal for "eventually". And if I can help make that> happen, I'll try to pitch in however I can. I have some experiments> I'm doing now, working on some new approaches to scaling RDF> triplestores, so using the Wikidata data may be an interesting testbed> for that down the road.>> And on a related note - and apologies if this has been discussed to> death, but I haven't been on the list since the beginning - but I am> curious if there is any formal collaboration> (in-place|proposed|possible) between dbpedia and wikidata?
Help with this would be awesome and totally welcome. The tracking bug is at https://bugzilla.wikimedia.org/show_bug.cgi?id=48143
Lydia,
Linked Open Data URIs for tracking issues such as the one above:
[1] http://linkeddata.uriburner.com/about/id/entity/https/bugzilla.wikimedia.org... [2] http://bit.ly/vapour-report-sample-wikidata-issue-tracking-entity-http-uri -- vapour report on the Linked Data URI above [3] http://linkeddata.uriburner.com/c/9BTVWIGG -- use of #this to make a Linked Open Data URI "on the fly" (no owl:sameAs reasoning and inference applied) [4] http://linkeddata.uriburner.com/c/8GUIAJ -- ditto, but with owl:sameAs reasoning and inference applied.
Since this mailing list is online, I can also add some RDF statements into this post. Basically, this turns said post (or any other such conversation) into a live Linked Open Data creation and publication mechanism, by way of nanotation [1].
## Nanotation Start ##
http://linkeddata.uriburner.com/about/id/entity/https/bugzilla.wikimedia.org/show_bug.cgi?id=48143 http://linkeddata.uriburner.com/about/id/entity/https/bugzilla.wikimedia.org/show_bug.cgi?id=48143 xhv:related https://twitter.com/hashtag/RDF#this https://twitter.com/hashtag/RDF#this ; is foaf:primaryTopic of http://linkeddata.uriburner.com/c/8GUHZ7 http://linkeddata.uriburner.com/c/8GUHZ7, http://bit.ly/vapour-report-sample-wikidata-issue-tracking-entity-http-uri http://bit.ly/vapour-report-sample-wikidata-issue-tracking-entity-http-uri .
## Nanotation End ##
Links:
[1] http://kidehen.blogspot.com/2014/07/nanotation.html -- Nanotation [2] http://linkeddata.uriburner.com/about/html/%7Burl-of-this-reply-once-its-liv... -- URL pattern that will show the effects (refied statements/claims amongst other things) of the nanotations above .
-- Regards,
Kingsley Idehen Founder & CEO OpenLink Software Company Web: http://www.openlinksw.com Personal Weblog 1: http://kidehen.blogspot.com Personal Weblog 2: http://www.openlinksw.com/blog/~kidehen Twitter Profile: https://twitter.com/kidehen Google+ Profile: https://plus.google.com/+KingsleyIdehen/about LinkedIn Profile: http://www.linkedin.com/in/kidehen Personal WebID: http://kingsley.idehen.net/dataspace/person/kidehen#this
Wikidata-l mailing list Wikidata-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata-l