[Talk-ko] Fixing laguage-mixed name tag in Korean region

[Max]

On 2017년 03월 04일 08:35, 느림보 wrote:
> I opened new thread to discuss automatic modification of name tag.

> Many POIs in Korean region have their name tag in /한국어(English)/form.
> Rule was changed but there was no action for modifying legacy names. It
> was the biggest harassment when I joined the community: “There are two
> naming rules. /한국어(English)/form is using without any documented
> guideline. Which style should I follow? Can I modify /한국어(English)
> /form to /한국어/form?”

I wouldn't say that there was no action. New items since have been added 
in the new scheme and some of the old ones have been modified to match 
the new scheme.

I also believe the process is kind of slow and unsatisfying. May be some 
kind of (semi-)automated process of conversion would indeed be good.

That said, I think there are valid objections to it too.
1. From what I have been seeing in the main talk list there is a great 
opposition against automated edits, and a strong belief in the crowd 
approach.
2. Many POIs in Korea are outdated and not valid any more. In the 
problematic import, there are all kinds of hospitals and clinics that 
are closed by now. Editing those would "mark them like new" and it would 
not be visible immedately that they might not be valid any more
3. There are other related issues that should be discussed and adressed 
too and possibly solved at the same time.

The current scheme is not perfect either. There are some issues to be 
discussed.

name=한국어
name:ko=한국어
name:en=Roman-ized
name:ko_rm=romanized

name=한국어 and name:ko=한국어 are kind of redudnant, but it is probably 
neccessary to help transitioning. Also it is the same way in Japan.

name:en=Roman-ized and name:ko_rm=romanized ok, I made a little joke, 
because in 80% of the cases they are identical and they should not.
That something is written in Latin characters doesn't make it English. 
Even if you capitalize it and write it in several words.
In my opinion the field en thould not be there if it isn't English. I 
regulary delete this field if it is identical to ko_rm

There are some elements like country, city or some subwaystations that 
have their name written in all kinds of languages in the OSM database, 
but this practice is questionable, It might be handled better in 
something like wikidata.

ko_rm should actually be renamed in bulk to ko-Latn, possibly in 
cooperation and discussion with the japanese community who have the same 
problem with ja_rm that should be ja-Latn

See here: https://wiki.openstreetmap.org/wiki/Names#Localization
the next paragraph in this wiki page is interesting too. We should avoid 
transliterations. According to this rule 90% of the name:ko_rm and 
name:en tags should go.

My opersonal rule of thumb is that I add the name:en tag only if the 
romanization could not be done automatically, or the original name is 
indeed English (most of the time it's both actually) like a Starbucks 
for example.

Most of the time the romanization follows a strict rule, meaning it can 
be done much better and accurately if done by the renderer and doesn't 
need to swell the database.

TL;DR
I agree, there should be some automatic or maybe semi-automatic process 
to clean up the map/database, but we need to discuss the issues here 
before and maybe even drag this discussion to the main email list to get 
a feedback..

m.