On Thu, May 8, 2014 at 9:14 AM, Wil Sinclair wllm@wllm.com wrote:
In highly structured databases, adding properties that may be useful for your research and the work of others would require altering the structure itself, like adding a field, for example. That isn't easy, because the powers that be have to first agree that it is appropriate, worthy, and fits correctly in to the ontology. If there is a type hierarchy, then the sample set would probably have to conform to a sub-tree in the type hierarchy which may not correspond well to the sample set that the researcher is actually interested in.
You can create your own instance of Wikibase and decide on the structure, fields, ontology, etc Then you can find the points of intersection with Wikidata concepts and link back to them if you feel like (not necessary, though). More info: https://www.mediawiki.org/wiki/Wikibase
I'm talking about the exact opposite, actually. Unstructured databases can be easily altered and indexed in much more flexible ways. The indices for these databases wouldn't normally be stored with the data itself; the researchers would get a data dump and create the indices needed for their own studies. Conventions would be enforced if the researchers wish to contribute anything back.
So basically like running an instance of CKAN? http://ckan.org/
Most importantly, if I understand correctly, wikidata is a secondary database that doesn't correspond one-to-one with Wikipedia articles yet, and it's not clear to me whether it ever will. While it might be interesting to someone using the data collected in Wikipedia and imported in wikidata for semantic-oriented research like basic AI that would help computers win on Jeopardy, it wouldn't be interesting to someone studying Wikipedia itself.
It might help to improve the data accuracy since it will be possible to update all uses of any parameter in any article at once. Some wikipedias use that data to generate text, I guess in those cases you could say that with quality data you will have quality text.
Cheers, Micru