On Wednesday 20 September 2017, Frederik Ramm wrote:
> Does that mean that the original plan of "fetching" (dare I say
> importing) 200k Wikidata links through automated connections from
> existing Wikipedia links is (a) dangerous because you can easily
> obtain a reference to something totally different, or (b) no problem
> because it is *to be expected* that an object's Wikipedia and
> Wikidata links point to different things an hence the import wouldn't
> introduce "errors" per se?

It all depends on how you use the data.  I think adding the wikidata 
tags is fine *because* i regard them as simple references to related 
features but if you'd insist on the idea that the OSM feature and the 
wikidata item refer to the same real world feature then inferring such 
an identity from an existing wikipedia tag is even more problematic - 
because the wikipedia tag was almost certainly not originally verified 
to refer to exactly the same concept as the OSM object.

Also keep in mind that both the OSM features and the wikidata items 
evolve over time and not every edit made in OSM (like extending the 
area of a forest polygon to include some additional tree covered area) 
is necessarily verified to still justify having the wikidata reference.

What will inevitably happen if you automatically add wikidata tags is 
that existing errors in either OSM (in form of incorrect wikipedia 
tags) or in wikidata (in form of incorrect connections to wikipedia 
articles) will get duplicated.

Christoph Hormann

talk mailing list

Reply via email to