[Wikimedia-l] Experience implementing Wikidata

2013-11-03 Thread Anders Wennersten
Since August we are implementing Wikidata for the Swedish 290 communes 
and 1900 towns, including automatic update from the authority 
Statistic Sweden to Wikidata.


Our original assumption was that it would take us 8 month but it now 
seems it will not be enough, it will probably be more like a year to get 
it fully implemented, and I thought our experiences could be of interest 
to several (even excluding anything of the WD community or the science 
of properites etc)


QUALITY
My assumption was that the Swedish articles was 98% correct but that 
Wikidata ought to have 99,8% correct info. We found, though, as a start 
that WD was only 96% correct as it has been loaded from en:wp, which is, 
from a Q point for these entities, worse then de.wp, nl:wp and sw:wp. A 
critical information, that was missing, was the communcode/towncode, 
that is needed to implement an automatic update. In practice this has 
meat we have had to manually check and complement all objects on WD, 
that with its primitive interface meant it taking a very long time, one 
of us even temporary getting repetitive strain injury


INTRICACIES
When we have created articles on sv:wp we have only covered some finer 
intricacies of these entities in the article text. But with Wikidata it 
is not enough, it has to be handled fully correct. Trosa commun existed 
1971-1973, was then merged into Nyköpings commun and 1994 broken out  
with some more areas to become again a Trosa commun. Sv.wp hase only one 
article  for Trosa commun, but for wikidata we have created two entities 
one called Trosa commun (1971-73), whos corresponding entity on sw.wp is 
a redirect. Another example: Statistic Sweden defines the boundary of a 
town by strange advanced citerias meaning a real town is some year  is 
seen by them as two towns. We have of these reasons been forced to 
define new properties, and create some hundred new entities in Wikidata, 
and check the corresponding redirect articles on sv:wp, giving them 
Categories etc, all OK but it takes time...


POTENTIAL
We are still convinced of the strong potential of WIkidata, but believe 
the full benefits requires an a (semi)automatic update of data from the 
authorities databases.  In our case Statistic Sweden are positive, 
believe in Open data and have a nice API, but it is a new component for 
us to take care of, the datatransfer/update program/script, who will be 
responsible for this etc?


Experience from others are welcome to us

Anders






___
Wikimedia-l mailing list
Wikimedia-l@lists.wikimedia.org
Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, 
mailto:wikimedia-l-requ...@lists.wikimedia.org?subject=unsubscribe

Re: [Wikimedia-l] Experience implementing Wikidata

2013-11-03 Thread Gerard Meijssen
Hoi,
Are you going to use this info in a blogpost or something ? If not is it ok
for me to do so ??
Thanks,
 Gerard


On 3 November 2013 09:02, Anders Wennersten m...@anderswennersten.sewrote:

 Since August we are implementing Wikidata for the Swedish 290 communes and
 1900 towns, including automatic update from the authority Statistic
 Sweden to Wikidata.

 Our original assumption was that it would take us 8 month but it now seems
 it will not be enough, it will probably be more like a year to get it fully
 implemented, and I thought our experiences could be of interest to several
 (even excluding anything of the WD community or the science of properites
 etc)

 QUALITY
 My assumption was that the Swedish articles was 98% correct but that
 Wikidata ought to have 99,8% correct info. We found, though, as a start
 that WD was only 96% correct as it has been loaded from en:wp, which is,
 from a Q point for these entities, worse then de.wp, nl:wp and sw:wp. A
 critical information, that was missing, was the communcode/towncode, that
 is needed to implement an automatic update. In practice this has meat we
 have had to manually check and complement all objects on WD, that with its
 primitive interface meant it taking a very long time, one of us even
 temporary getting repetitive strain injury

 INTRICACIES
 When we have created articles on sv:wp we have only covered some finer
 intricacies of these entities in the article text. But with Wikidata it is
 not enough, it has to be handled fully correct. Trosa commun existed
 1971-1973, was then merged into Nyköpings commun and 1994 broken out  with
 some more areas to become again a Trosa commun. Sv.wp hase only one article
  for Trosa commun, but for wikidata we have created two entities one called
 Trosa commun (1971-73), whos corresponding entity on sw.wp is a redirect.
 Another example: Statistic Sweden defines the boundary of a town by strange
 advanced citerias meaning a real town is some year  is seen by them as two
 towns. We have of these reasons been forced to define new properties, and
 create some hundred new entities in Wikidata, and check the corresponding
 redirect articles on sv:wp, giving them Categories etc, all OK but it takes
 time...

 POTENTIAL
 We are still convinced of the strong potential of WIkidata, but believe
 the full benefits requires an a (semi)automatic update of data from the
 authorities databases.  In our case Statistic Sweden are positive,
 believe in Open data and have a nice API, but it is a new component for us
 to take care of, the datatransfer/update program/script, who will be
 responsible for this etc?

 Experience from others are welcome to us

 Anders






 ___
 Wikimedia-l mailing list
 Wikimedia-l@lists.wikimedia.org
 Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l,
 mailto:wikimedia-l-requ...@lists.wikimedia.org?subject=unsubscribe
___
Wikimedia-l mailing list
Wikimedia-l@lists.wikimedia.org
Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, 
mailto:wikimedia-l-requ...@lists.wikimedia.org?subject=unsubscribe