[OSM-talk] Adding Wikidata tags to 70k items automatically

[Janko Mihelić]

Hi Frederik, great post.

I always thought wikidata tags would be a first step to a new database that
OSM needs, and which would be something like Wikidata is to Wikipedia. A
database that classifies and categorizes elements in OSM. In my opinion a
slightly modified version of Wikibase could be enough. It would have
combinations of tags and areas where that category applies.

If we had a database item in our potential OSM-wikibase that said "all
primary highways with ref 55 in the UK" and then offload the proprietary
identifiers there, I think no one would mind.

Another problem this potential database solves is the often quoted
"relations aren't categories". We could make an item "HSBC ATM machines"
and add a few tags that define an element as a HSBC ATM machine
(amenity=atm + operator=HSBC, or amenity=atm + ref:HSBC=34 or whatever)

Even wikidata tags that we are discussing could be deprecated if we had
such a database. We would have our own identifiers (O3489 instead of
Q23489) and those identifiers would link to wikidata as well as to other
databases.

If we don't make our own database that categorizes our tags and elements,
someone else will do it. In this case, it's Wikidata, which is better then
the rest because it is crowdsourced, it has an open license for it's data,
and it has opensource software for it's database and user interface. So if
I'm not mistaken, we could fork their project if that was needed. But if I
could choose, we would have our own.

Janko Mihelić

2014-08-30 23:40 GMT+02:00 Frederik Ramm <frederik at remote.org>:

> Hi,
>
> On 08/29/2014 08:10 PM, Rob Nickerson wrote:
> > Lets just step back and reflect on this for a minute.
>
> My overarching concern is that if this import is done, future imports
> will use the "but we also have Wikidata links" argument for justification.
>
> What we have here is a third-party database whose object identifiers we
> add to OSM as tags in order to make linking things easier.
>
> This is something that has often been requested by people but never been
> granted on a large scale because we always said that it would be an
> abuse of our database and our mapper's patience to offload everyone's
> and their dog's linking requirements onto us.
>
> Imagine every single stretch of road being tagged with three different
> proprietary (or at least competing) identifiers of traffic information
> providers, or similar payload tacked onto OSM.
>
> I'm not saying no to a Wikidata import but I would like the proponents
> to state clearly what makes this import special - why this and not, for
> example, Ordnance Survey TOIDs or IDs of Mapillary photos?
>
> I would like to define high hurdles for such an import. The import costs
> us a lot (in terms of mappers having to spend time to understand the
> tags and know how to handle them when they edit an object, more quality
> checks, etc.), and it must be proven - or at least expected - to offset
> that cost by making itself useful. It's ok of this particular import
> clears the hurdles but at least I can then use the hurdles if the next
> guy in line comes along and wants to import his keys too.
>
> Also, I note that Rob wrote:
>
> "We have our own restrictions (verifiable on the ground, not changing to
> frequently) in OSM. A link to wikidata allows us to continue with these
> restrictions but still allow people to get at interesting non-geographic
> data."
>
> And Andrew said:
>
> "A few more points: This isn’t a deletionists’ charter and we shouldn’t
> rush to unload any tagging onto Wikidata without discussing the removal
> very carefully."
>
> I am personally very much in favour of "unloading" non-geo-database
> tagging elsewhere. We started with marking restaurants (useful) and
> recording their names (also useful), classing them into fast-food or
> "proper" restaurants, then tagging what cuisine they have and how to
> contact them for booking a table, and meanwhile we're recording whether
> they have vegan food and what their opening times are and whetehr you
> can pay with bitcoin. We don't currently have anywhere to "offload" that
> information which is all useful for certain use cases, but once we have
> a working integration, I should very much hope that stuff like the menu
> and the opening times will be recorded either in OSM or in Wikidata but
> not both.
>
> Allow me one question in that matter as a Wikidata ignorant though: Are
> there any notability rules in Wikidata? For example, if we should one
> day decide that restaurant opening times should not be recorded in OSM
> but on Wikidata, can it then happen that someone records a restaurant's
> opening times on Wikidata only to have that information kicked out by
> someone else again for lack of relevance?
>
> Because if that were the case, then Wikidata would be relatively useless
> as a general offloading point for non-geodata.
>
> Bye
> Frederik
>
> --
> Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"
>
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20140831/df5d8669/attachment.html>