Some Japanese place names are transliterated to their alphabetical form using Chinese phonetic readings.

[stuartcw]

I'm scrolling around the town where I live in Japan on https://openfreemap.org and I notice that many Japanese Place names have been transliterated into alphabetical strings in the way that they would be read in Chinese.

It appears that the phonetic reading generation is being done by some automatic process which is selecting the Chinese way of transliterating them instead of the Japanese.

This is a very tricky to get right as there is no foolproof way of obtaining the phonetic reading from Japanese Kanji Characters. (Hiragana and Katakana phonetic scripts are trivial.)

Therefore, a place name in one part of Japan has one reading yet the identical place name in another part of Japan has a different phonetic reading. Even Japanese readers have to check to be sure how to read a place name that they are unfamiliar with.

Difficult place names are quite rare but mean that you have to look up whether a place name is an exception and use the correct version before using a dictionary of the common transliterations.

The current Chinese transliteration of Japanese placenames is not useful at all, actually harmful, so I would suggest turning it off until a better solution or fix is found.

[hyperknot]

Hi @stuartcw and thanks for your report.

Your insight is valuable because I have no idea about Japanese place names, nor does probably most OSM developers.

As a first step, can you find out if there is any language in the dataset with correct names?

You can do this by going to Maputnik / View / Inspect here: https://maputnik.github.io/editor?style=https://tiles.openfreemap.org/styles/bright

Then when you click on one of the little red dots you'll see the names in all possible language. Does any of them look good to you?

[hyperknot]

Like this:

[stuartcw]

I have had a quick look and it seems that there is a mix of correct and incorrect data. i.e. for the subclass:bus_stop the name:latin is Chinese. Whereas subclass:station the name:latin is correctly transliterated.

For the incorrect subclasses it seems like the int_name (internationally recognized name) is the Chinese transliteration and this is mirrored to the name:latin. In the more prominent features like example above (prefecture name) all languages seem OK. I think the problem is limited to the more local labels such as bus_stop, pub etc.

I can look through the other subclasses later and check them if you like.

[hyperknot]

I think it's probably changes on an item-by-item basis. The important ones are city names, etc. So the big question is if there is a data transformation issue between OpenStreetMap -> OpenMapTiles or if the problem is in OpenStreetMap itself.

For example for Iwate Prefecture, I found this in OSM: https://www.openstreetmap.org/relation/3792412 and also this: https://nominatim.openstreetmap.org/ui/details.html?osmtype=R&osmid=3792412&class=boundary

[hyperknot]

Wrote a document for debugging international names: https://github.com/hyperknot/openfreemap/blob/main/docs/debugging_names.md

[stuartcw]

Here’s one example. Also all the minor features are in the area are the same. The problem is that the displayed is "jiǔ lǐ bāngF・marinosu tōngri" which is the Chinese rendering of the name. This has no meaning in Japanese or English. When I use OpenStreetMaps I see "Kurihama F. Marinos Dori" as expected. I only see the the Chinese name “ jiǔ lǐ bāngF・marinosu tōngri ” when inspecting maputnik or viewing openfreemap.org

久里浜F・マリノス通り

https://www.openstreetmap.org/way/159187467#map=18/35.232067/139.701945

https://nominatim.openstreetmap.org/ui/details.html?osmtype=W&osmid=159187467&class=highway

https://maputnik.github.io/editor/#17.25/35.232219/139.702426

1591874672

name_int: jiǔ lǐ bāngF・marinosu tōngri name_latin: jiǔ lǐ bāngF・marinosu tōngri

[hyperknot]

Thanks, I'm trying to understand the issue step-by-step.

1. So we are talking about this label here:

2. In the default style, here is how it's constructed:

So first it's name_latin and then name_nonlatin.

3. What you are saying is that name_latin is wrong and shouldn't be displayed, right?

Now, what I cannot see correctly, is name:nonlatin, name_de, name and name_en all correct? The non_latin has one more character in the right side, is that good or bad?

[1ec5]

As reported in https://github.com/onthegomap/planetiler/issues/86#issuecomment-1049998723, name:latin is all but useless for romanized map labels.

[j9d3it]

I've found that this Python library is really great for romanization of place names. https://github.com/polm/cutlet

[curbengh]

Now, what I cannot see correctly, is name:nonlatin, name_de, name and name_en all correct? The non_latin has one more character in the right side, is that good or bad?

the screenshot doesn't show the full text.

The original values:

metadata key	value
name_int	jiǔ lǐ bāngF・marinosu tōngri
name:nonlatin	久里浜・マリノス通り
name_de	久里浜F・マリノス通り
name	久里浜F・マリノス通り
name:latin	jiǔ lǐ bāngF・marinosu tōngri
name_en	久里浜F・マリノス通り

My attempt:

metadata key	value	comment
name_int	Kurihama F. Marinos Avenue
name:nonlatin	久里浜・マリノス通り	Not sure what's the use of this key, name without latin script? "F" is removed.
name_de	Kurihama F. Marinos Avenue	German?
name	久里浜F・マリノス通り	Original name
name:latin	Kurihama F・Marinosu Dōri
name_en	Kurihama F. Marinos Avenue

Anyhow, I think this is more of an upstream issue. The issue of Chinese transliteration for the Japanese words (Kanji) is also mentioned in the OSM wiki.

---

I'm switching to Protomaps as a workaround as it offers language-specific style (a feature that is also nice to have here #22). In that style, if a POI doesn't have english name, then the label only has local name.

Protomaps

OFM