[OSM-talk] Overlapping brands (was "Fixing wiki* -> brand:wiki*")

Martin Koppenhoefer dieterdreist at gmail.com
Wed Sep 27 22:45:25 UTC 2017



sent from a phone

> On 27. Sep 2017, at 23:09, Yuri Astrakhan <yuriastrakhan at gmail.com> wrote:
> 
> Martin, that specific Wikidata item may have some, possibly incomplete data, that can be easily fixed, but that's irrelevant. As I keep saying - the wikidata and wikipedia tags are no different - both point to the same article in Wikipedia.


I think there’s a fundamental difference: when I link a wikipedia article I will take a look at the specific article in the specific language and then link it. When I link a wikidata item I will look at the properties of this item and then link it. I will not read all linked wikipedia articles in all languages, and from my experience (looking only at 3 languages) the wikipedia articles in different languages differ _a lot_. There are much more often than you‘d probably expect articles dealing with different things interlinked. And articles dealing with somehow the same subject still differ a lot, in all regards, what they cover, what is included (and what is in a different article), etc.

Yes, most wikipedia articles point to wikidata objects and most WD obj point to WP articles, but this doesn’t mean you can blindly assume they are telling the same story/containing the same content.

The stability issues exist in Wikidata as they do in Wikipedia, just on a different level: wikipedia articles might get merged, renamed or deleted (and you can mostly see this because your link will be dead), but wikidata objects change as well (classes (instance of) change, new properties get added, others are removed or their meaning changes, etc.)

Yes, it might be possible to fix (sync) everything, but there aren’t enough people actually doing it, and it is much easier to create hard to detect new problems by changing existing items „with a click“. And because everything is linked, it is also impossible to overlook the incredible complexity (unlike wikipedia, where an article can be modified a lot but still will tell a similar story or the edit will typically get reverted, or it was wrong before).

Maybe it’s really and mostly just an issue of too few people verifying and editing Wikidata (e.g. if more properties and other information would be added to a WD object, people could easier recognize when the link of the WP article in their language is pointing to the wrong object, but it just doesn’t seem to happen in reality: WD said for years that human settlements were administrative territorial entities, and it was only corrected after it was mentioned here.
It still says a human settlement is a subclass of a geographical point, looking at the geographical point object I learnt this is: „point or an area on the Earth's surface or elsewhere“ (btw: it’s also a „part of“ (object of which the subject is a part.) Earth’s surface, which seems inconsistent as „elsewhere“ is not part of Earth‘s surface). Now the settlement is at the same time („all instances of these items are instances of those items“) also a subclass of a „geographic region“, but the region is „different from“ („item that is different from another item, but they are often confused
is notnot to be confused withdistinct fromnot the same asnotisn'tdistinguished fromdifferent thansee alsodisambiguated fromnot same as“) a geographical point.

Which contradictions can be ignored and which should be fixed?

cheers,
Martin 





More information about the talk mailing list