Hi all,


thank you for the interest in the primary sources tool!


I wanted to make sure that there are no false expectations. Google has committed to deliver the initial tool. Thanks to Thomas P’s internship and support from Thomas S and Sebastian, and with the release of the data, the code, the paper, and all services running on Wikimedia infrastructure, we have achieved that milestone. The tool was developed as open source, in order to allow the community to continue to mold it and to invest in it as the community sees warranted.


I am particularly thankful to Marco Fossati for his work in creating further datasets. Thomas S has in the last few days cleaned up the issue list and merged pull requests. Thank you, Thomas! We are all very thankful for the pull requests, in particular to Thomas P, Wieland Hoffmann, and Tom Morris. In general, we plan to keep the tool up as far as our time allows, and continue to merge such requests, but we have no concrete plans of extending its functionality right now.


We are very grateful to everyone contributing to the project, or using the tool. If anyone wants to take over the project, we would invite you to contribute a bit for a while, and then let’s discuss about it. I would be thrilled to see this tool develop.


As a reminder, a lot of data has been released under CC0. We invite all to play around with the data and see if there are slices of the data that can be directly uploaded to Wikidata, as Gerard suggests.


If there are any questions, we’ll try to answer them. Again, thanks everyone!


Cheers,

Denny

On Tue, Feb 23, 2016 at 11:08 AM Tom Morris <tfmorris@gmail.com> wrote:
On Tue, Feb 23, 2016 at 1:52 PM, Stas Malyshev <smalyshev@wikimedia.org> wrote:

> As Gerard has pointed out before, he prefers to re-enter statements
> instead of approving them. This means that the real number of "imported"
> statements is higher than what is shown in the dashboard (how much so
> depends on how many statements Gerard and others with this approach have
> added). It seems that one should rather analyse the number of statements

Yes, I do that sometimes too - if there is a statement saying "spouse:
X" on wikidata, and statement in Freebase saying the same but with the
start date, or the Freebase one has more precise date than the Wikidata
one, such as full date instead of just year, I will modify the original
statement and reject the Freebase one.

I filed a bug report for this yesterday: https://github.com/google/primarysources/issues/73
I'll add the information about more precise qualifiers, since I didn't address that part.
 
I'm not sure this is the best
practice with regard to tracking numbers but it's easiest and even if my
personal numbers do not matter too much I imagine other people do this
too. So rejection does not really mean the data was not entered - it may
mean it was entered in a different way. Sometimes also while the data is
already there, the reference is not, so the reference gets added.

Even if you don't care about your personal numbers, I'd argue that not being able to track the quality of data sources feeding the Primary Sources tool is an issue.  It's valuable to not only measure quality for entire data sets, but also for particular slices of them since data sources, at least large ones like Freebase, are rarely homogenous in quality.

It's also clearly an issue that the tool is so awkward that people are working around it instead of having it help them.

Tom 
_______________________________________________
Wikidata mailing list
Wikidata@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata