[OSM-talk] iD news: v1.9.6 released

Minh Nguyen minh at nguyen.cincinnati.oh.us
Sat Jun 18 01:44:14 UTC 2016


(Sorry, I sent from the wrong address, so this message got stuck in moderation.)

Martin Koppenhoefer <dieterdreist <at> gmail.com> writes:

> 
> 
> 
> Sorry that it took me a while to reply, I was out of office and just came
back now.
> 
> 
> Il giorno 10 giu 2016, alle ore 21:07, Minh Nguyen <minh <at>
nguyen.cincinnati.oh.us> ha scritto:
> 
> 
> 
> is this your interpretation or is it explicitly defined like this? I'm
> astonished that these 2 concepts are supposedly structured vertically and
not horizontally in wikidata 
> 
> 
> <https://www.wikidata.org/wiki/Q486972> is defined in Wikidata as a
subsetof <https://www.wikidata.org/wiki/Q56061>. I agree that this is a
suboptimalrelationship – anyone can edit
> 
> 
> 
> it's not a big problem to have errors in the relations as long as
everybody can agree that it's an error and should be corrected. But the fact
that this wrong relation still sits in there and doesn't actually get
corrected is a bit troubling, particularly as this is IMHO a major bug which
has influence on all settlements that are in wikidata.
> 
> 
> Another issue I believe to have found in this item looking at
https://www.wikidata.org/wiki/Q486972 :the first words, (I believe it is
meant to be the definition, although it does not explicitly say so,) are
"densely populated geographic location
> 
> 
> populated place
> settlement
> human community
> inhabited place"
> 
> Now when you think about a ghost town, it clearly is a human settlement,
but this definition falls short in covering it. I would prefer, rather than
a collection of related terms (human community, inhabited place etc), a
complete sentence that defines the term, e.g. "a human settlement is a place
where a community of humans lives or lived and decided in the past to settle
and create dwellings." (surely could be improved, just an example).
> 
> There are lots of unanswered questions in wikidata, and probably, by
judging from the current state of the data, not enough editors to manage to
look through all of it.
> Another issue that comes to mind: what about contradictions between an
article and a wikidata object? How is the relationship between the articles
and wikidata? Apparently, you can only associate on article to one item, so
this suggests a strong relationsship, but clearly there are articles in many
languages, and there will be lots of contradictions between those languages
(because there are lots of articles, and because nearly noone cross checks
different languages). IMHO it would have been a better idea to have a less
strong relationship, something like "this article has information about this
item" (and hence allow multiple articles to be associated with an object)
rather than what it seems conceptually  to be thought of now (the articles
and this wikidata object deal with the same thing, are the same thing, here
in article form and here as mathematical relations).For me these are just
more indications that we should not base automatic edits on this source at
the moment. 

You raise a number of important points, but I think these concerns should be
posed on Wikidata's village pump or mailing list, where they can be more
effectively addressed. The Wikidata community has more answers that I would
individually.

I disagree that these considerations make the iD feature less correct given
that the wikidata tag is already widely used.

In fact, many of these same questions could be asked of Wikipedia. If the
mapper is tagging a ghost town POI with a Wikipedia article, can we be sure
at a technical level that the article's contents or categorization matches
OSM's semantics? Regardless, if the tagged feature is tagged as a ghost town
in OSM, the fact that it is tagged as a non-former human settlement in
Wikidata merely means the Wikidata item needs to be retagged as a ghost
town. I see no problem with the semantics in
<https://www.wikidata.org/wiki/Q74047>.

When an English-speaking mapper tags `wikipedia=en:China`, what's to say
`wikipedia=fr:Chine` has the same semantics? It's two hops away from the
English article, rather than the one hop to Wikidata. Yet most tools and
users assume that one is equivalent to the other when tagged on an object.
Otherwise, we'd have to maintain a litany of redundant `wikipedia:xy` tags
on every place POI and basically try to replicate Wikidata's interwiki
function for any Wikipedia tag to be broadly useful.

It wasn't so long ago that most Wikipedias' articles named "Crimea" referred
to the political entity as well as or instead of the peninsula. If, back
then, the administrative boundary in OSM had been tagged with
`wikipedia=en:Crimea` and mappers subsequently forgot about that tag, it
would've suddenly become illogical as soon as Wikipedia editors decided to
move the "Crimea" article somewhere else and replace it with what was
previously titled "Crimean peninsula". (The opposite happened with the
"China" article.) Meanwhile, the `wikidata` tag would've continued to point
to an administrative territorial entity item.

I'm making these arguments under the assumption that the `wikipedia` and
`wikidata` tags are acceptable in their own right. If you disagree with that
assumption, that's a broader discussion than whether iD should be
associating one with the other in response to a user selection.

> Generally, in OSM we are proud about the high quality of our data, because
the mappers base their work on first hand experience and human judgement,
and we shouldn't give these up just to gain a little more editing
convenience (IMHO).
> 
> We do have very strict guidelines for imports and automated edits, and
IMHO this new features falls into the second category...

I'm familiar with our guidelines on automatic edits, but I find it a stretch
to apply them here. iD is essentially inserting an alternative, more stable
representation of a title the mapper specifically chose. This is no more
automatic than an editor silently deleting TIGER tags upon touching a node,
filling in the `source` tag of a changeset based on the imagery layers used,
or stripping spaces after semicolons in tags.

I agree that editors should respect their users' judgment. But many of us
chose an editor based on how productive it makes us. I personally prefer iD
over JOSM because of iD's intuitive UI (IMHO) and despite the lack of other
conveniences like 3D editing and find and replace. Someone might prefer an
editor on their phone in order to map with one hand while walking about,
even though they may lose precise control over tagging. In any case, the
`wikidata` tag can be edited or deleted manually in iD by any user who
prefers to omit it.

-- 
Minh Nguyen <minh at nguyen.cincinnati.oh.us>


More information about the talk mailing list