[OSM-talk] Wikipedia/Wikidata admins cleanup

Andy Townsend ajt1047 at gmail.com
Wed Jan 4 13:09:04 UTC 2017


On 04/01/2017 12:19, Andy Mabbett wrote:
> On 4 January 2017 at 10:11, Frederik Ramm <frederik at remote.org> wrote:
>
>> Wikidata tags in OSM are already of low quality because of mindless
>> mass-addition by people with zero local knowledge
> Oh, please stop with this FUD.
>
> Unless you have evidence* to support this assertion, you should
> apologise to the good people whose hard work has seen many thousands
> of Wikidata tags usefully - and automatically - added to OSM.
>
>
> * not anecdotes.
>
>
>

Do my comments on http://www.openstreetmap.org/changeset/43749373 count 
as "evidence" or "anecdotes" in your book?  I'm still waiting for the 
issues raised there to be addressed.

FWIW (and this bit is anecdotal - I've not put any numbers together) the 
root cause of the problems I'm seeing local to me seems to be a lack of 
data quality at the wikidata end. Typically, a wikidata article is 
created from a wikipedia article, and that wikipedia article covers more 
than one OSM item.

For example, https://www.openstreetmap.org/relation/957808 is Dunham on 
Trent civil parish, and https://www.openstreetmap.org/node/1782116865 is 
the village itself.  The civil parish object has a wikidata link to 
https://www.wikidata.org/wiki/Q5315240 , but the village does not.  
Unfortunately https://www.wikidata.org/wiki/Q5315240 claims to be a 
village not a civil parish.  It was created from 
https://en.wikipedia.org/wiki/en:Dunham-on-Trent,%20Nottinghamshire 
which claims to be _both_.

You could argue whether the problem occurred at the creation of the 
wikidata item from the wikipedia item, or the linking of the "wrong" OSM 
item to wikidata, but you can't argue that something hasn't gone a bit 
wrong here.

I'd certainly suggest that currently anyone consuming wikidata links in 
OSM data looks very carefully at who added them.  For example, when I 
create Garmin maps (mainly for walking) I try and indicate on them if a 
contributor isn't likely to have actually walked the path in question - 
if I need to get back to a location before it gets dark I might choose a 
path added by someone who's definitely walked down it rather than 
someone who may not have done - maybe wikidata consumers should do 
something similar.

Best Regards,

Andy (SomeoneElse)





More information about the talk mailing list