[OSM-talk] Can wikidata links help fight name inflation?

moltonel 3x Combo moltonel at gmail.com
Thu May 28 12:20:20 UTC 2015


On 28/05/2015, Frederik Ramm <frederik at remote.org> wrote:
> It is my impression that a large proportion of name:xx tags in OSM are
> added by "naming specialists" who do little else than large scale name
> additions.

Nothing wrong with that, a lot of OSM contributors specialize in some
type of data.

> it would probably not be too much to ask for them to indulge
> Wikidata instead of OSM.

Probably not.

> If we could offload 99.99% of all name:xx tags to
> Wikidata and keep them only in edge cases like your Scalinata di Trinità
> dei Monti, why not? Why would a few cases in which name:xx tags remain
> ruin the whole scheme?

I really like the idea of offloading "translation" work to wikidata,
but I see important issues :

1) Consuming the data at scale is not trivial. Querying wikidata API
for each osm object with a wikidata link would kill performance.
Consumers would need to mirror wikidata (with some luck just a dump of
Id->names) and keep it as up to date as the osm data.
1.2) It'd be a great thing to do regardless of what OSM decides about
its name:CC tags. But OSM shouldn't change its behavior until 1 or
more major rendering, geocoding, and routing projects have implemented
the workflow and come back to talk about it.

2) The cutoff for which name:CC belongs into OSM or not is arbitrary.
Ireland's first official language is Irish, but in most you won't find
anybody who speaks mainly Irish (yet we hope to reach 100% name:ga
coverage in OSM). At the other end of the scale, London is one of the
most cosmopolitain city in the world, so there's a fair chance that
all 154 London names in OSM correspond to the native language of
somebody currently living there.

3) Contributing to OSM is already very hard: most people in OSM's
target audience of "people whi have VGI to contribute" are not
tech-savvy. Raising the bar by making wikidata (with its different UI
and community style) doesn't sound like a good idea.



More information about the talk mailing list