[OSM-talk] Name:* tags in the local language

Paul Norman penorman at mac.com
Tue Apr 24 17:23:19 UTC 2018


As part of my Wikimedia Foundation work, I'm working on labeling in
multiple languages using OSM data. We've run into an issue, and it's not
clear how to best solve it.

It is sometimes recommended that when you add a name in another language
you also indicate the name in the local language by adding a suitable
name:* tag at the same time. For example, if adding "name:fr=Londres" to
London, you would also add "name:en=London" if it weren't present.

This practice is not widely followed. For example, 3% of the roads with
names in multiple languages in the western US have a name:en tag.[1]
Cities are better, but still only 30% of cities and towns with names in
multiple languages have a name:en tag. For the same criteria in China
and with Chinese, it's 30% of roads and 75% of cities.[2]

This is a problem for displaying labels. Proper display of labels is
more than just showing the name in the language the user requested and
falling back to the name=* tag. It requires falling back through
multiple languages. For example, if someone requests a map in
Luxembourgish, you would show German labels where places and objects
have no Luxembourgish names. There are similar situations for French
Creoles and many regional and minority languages.

We'd like to be able to do more. Specifically, we'd like to always show
labels in the user's script if possible. For example if a French user is
viewing a map of India, they're more likely to be able to read an
English name than one in a Dravidian language, Hindi, or Bengali. The
common lack of a name:* tag for the primary language while having other
name:* tags makes this impossible.

There aren't any great solutions. Fixing this in software doesn't work
well because it requires regional processing of what are increasingly
small regions as you get to less used languages. Language detection from
the tag value is fragile and introducing magic logic. This really needs
addressing on the OSM data side.

If there's agreement that there is a problem here, I could look at
preparing a mechanical edit or MapRoulette challenge to add name:* tags,
e.g. adding name:en to objects in the US with other name:* tags, and
adding name:zh in China. As an estimate, this would be 115k changes in
China, touching 28% of roads there.

[1]: https://phabricator.wikimedia.org/T192662#4151714
[2]: https://phabricator.wikimedia.org/T192662#4147291




More information about the talk mailing list