-
Notifications
You must be signed in to change notification settings - Fork 120
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
🌍 Low zoom locality names are missing translations + country, region zooms #977
Comments
We'd need to match the selection of features shown in earlier zooms when changing the source, but the extra names would sure be nice. |
Related: #1140 |
OpenMapTiles uses some nifty logic to merge city ranks between OpenStreetMap and Natural Earth, we could probably do similar for the names? (Keeping the Natural Earth features, but decorating them with additional OpenStreetMap names at runtime.) |
Might it make sense to do the inverse; keep the OpenStreetMap places, but decorate them with Natural Earth |
Copying @matkoniecz issue comment from #1371 (comment) here:
https://en.wikipedia.org/wiki/%C5%81%C3%B3d%C5%BA has label "Lódz" I would expect either Łódź (local name, name on English wikipedia) or Lodz in ASCII form. Not unholy contraption that is neither local nor English name, probably because NE has problem with UTF8.
English names without local names, as result map that is trying to show labels in specific language is broken on low zoom
OSM has several languages, and specifies proper local name.
In other words - I think that for cities it is desirable to discard NE and use pure OSM (or merge OSM with NE). https://github.com/gravitystorm/openstreetmap-carto has quite good (see https://www.openstreetmap.org/#map=5/51.272/21.583 or other areas on low zoom) display of labels that IIRC is based solely on OSM data (if useful I may track down and summarize how this is done)
Problem is global, systematic and affects all low zoom where OSM data is not used for city labels.
Encountered for example on https://mapzen.com/tangram/play/?#8.7500/48.1808/16.2415 (Vienna) during development of https://gist.github.com/matkoniecz/456041cea84e0cc58a82a070acd5b4b1 style (if problem is not clear I may produce minimal working example). |
@matkoniecz I agree low-zoom name localization is an important area that needs further work, and we're hoping to tackle it in the next several months. Generally speaking we used to only use Natural Earth at low-zooms and OpenStreetMap at mid- and high-zooms, but in v1.0 we started cross-fading between the two at around zoom 8, and some features like landuse and POIs are only from OpenStreetMap even at low-zooms. Some features like country labels and continent labels are just from OpenStreetMap at all zooms (and include the expected translations). Some of OSM and NaturalEarth mixing is for historical reasons, and some of it is for my own vanity as the primary author of Natural Earth – when there are problems with Natural Earth my tendency is to want to fix them instead of just switching over to only OpenStreetMap. One thing that is very nice from the Natural Earth side is knowing which few dozen to few hundred locality features to include at low zooms instead of stuffing the tiles with 1000s of OpenStreetMap features that can't all be labeled anyhow. This improves file size two fold by limiting the number of features included, and because each new low-zoom feature is highly likely to have lots of translations, limits unique (poorly compressed) name property values at low zooms. Limiting the number of features also improves rendering performance. (We also have too many OSM country labels with all their translations when they can't all label in the tiles now.) There are a few options on the table:
We've spent the summer investigating locality name translations in the low zooms (property values, not engineering options listed above) and will have a blog post up about that probably next week. And for your original issue about the misspelled placename: Looks like we goofed up |
based on my experience with fixing OpenMapTiles similar algorithm , I would prefer the WIKIDATA based merge/join solutions. so my suggestions: |
Adding wikidata codes to Natural Earth is feasible. I have a partial mapping in Who's On First now. Do you have a mapping to contribute, compare with?
… On Aug 17, 2017, at 19:04, ImreSamu ***@***.***> wrote:
There are a few options on the table:
Keep using Natural Earth features but join them with OSM at run-time to add in missing properties.
See comments above for how OpenMapTiles does it.
based on my experience with fixing OpenMapTiles similar algorithm , I would prefer the WIKIDATA based merge/join solutions.
so my suggestions:
6. Adding wikidata code for Natural Earth, and join with OSM via wikidata value. ( As I know Natural Earth does not contain wikidata information yet )
Maybe we can reuse the OpenMapTiles algorithm to generate wikidata code for Natural Earth database,
―
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or mute the thread.
|
I try to help, I need a little time to analyze the best solution
IMHO: The OSM licensing issues the biggest problem ( ODBL vs. Public domain )
Detecting/Checking:
Comment: |
Why? http://www.naturalearthdata.com/about/terms-of-use/ and wikidata are PD, so after mixing OSM with Wikidata and NE data license is not going to be a problem. One may simply use ODBL. |
Many of the wikidata concordances are already available via Who's On First under CC0 – I'll investigate adding them to Natural Earth from there, with verification QA steps against OSM to make sure the joins there result in expected behavior, and do one-off corrections: |
Tracking the Wikidata concordance work upstream in nvkelso/natural-earth-vector#214. |
+1 on bringing WikiData codes into the next Natural Earth release. It would be truly amazing @nvkelso |
It affects also water areas - for example Aral Sea ( http://www.openstreetmap.org/relation/2195612 ) has no Polish label on lower zoom levels. |
Usually |
I’ve been working on a Mapzen-based project — http://majes.tc/toponyms (background here) — that shows native-language place names, and then displays a choice of localized name (if available). Was excited to see the WOF work expanding the number of name translations as I want to provide a long list of potential localizations. This whole project is predicated on displaying local names for everything and I thought I’d chime in on this issue: The Bubble Wrap style on which I’ve based my map appears to use Natural Earth data for major city names (and some other features) at low zoom levels and these do not appear in their native language until they switch over to OSM at zoom level 8. Is there a way to workaround this, by forcing OSM/WOF as a source so that native place names are shown at all zoom levels? Unlike much of what you’re undoubtedly focused on, this is a side project with little or no commercial potential, but it’s one of my first map projects and I’ve been amazed at how easy Mapzen has made it to bang out this prototype, so thanks for that. Cartographic rookie, but let me know if there’s anything I can do to help? |
Hi @drewwilliams. Thanks for your feedback. We have some immediate improvements planned that will pick up ~8 languages for low zoom populated places from the new Natural Earth v4 release (Dec/Jan release). But that's only a stop gap until one of the other proposed solutions in #977 (comment) can be acted on. |
Since OSM region to NE regions have poor mapping, even using Wikidata ID joins, let's investigate default min_zoom, max_zoom per country (with a table @nvkelso will provide) using the roads intercut method and then when there is a match use the curated value instead. For the countries it seems to be good enough match rate we can fix up any Wikidata ID funk (like was done for Taiwan). |
Verified on dev. #858 Tracks adding WikipediaID later in v1.6 milestone. |
tl;dr
The new Natural Earth v4.1 populated places file now includes localized names in a
name_*
, specifically:dozen themes expanding from populated places to include all admin-0, admin-1,
rivers, lakes, playas, geographic lines, physical labels, parks, airports, ports, and
more. As part of this work a new unique and stable "ne_id" has been added for any
feature with a name translation &/or a Wikidata ID concordance. The full list of
languages is:
name_ar
,name_bn
,name_de
,name_en
,name_es
,name_fr
,name_el
,name_hi
,name_hu
,name_id
,name_it
,name_ja
,name_ko
,name_nl
,name_pl
,name_pt
,name_ru
,name_sv
,name_tr
,name_vi
, andname_zh
(Names with * indicate
new language in v4.1 series.) A 2-character language code decoder ring is here:
https://en.wikipedia.org/wiki/List_of_ISO_639-2_codes. Props to Wikidata for their
CC0 license: https://www.wikidata.org/wiki/Wikidata:Introduction.
This is for many themes, not just populated places. So work here should be generic for many theme imports.
===
Original title: Include OpenStreetMap language data for Natural Earth features
For example, ocean names are available in many languages at zooms 9 and higher, but only in English at zoom 8 and lower. Example map: https://mapzen.com/tangram/play/?scene=https%3A%2F%2Fapi.github.com%2Fgists%2F9f177672cdaac4640ea1b519cde3420d#9.4233/13.3585/-38.4174
If you zoom out below zoom 9, the label disappears.
The text was updated successfully, but these errors were encountered: