User:Frederik Ramm/When is a name a name?

From OpenStreetMap Wiki
Jump to navigation Jump to search

Although this is a page in my user space, I explicitly invite my colleagues from DWG to edit at will.


OpenStreetMap has a policy on naming in disputed areas [insert link], but outside of such disputes, the "name" and "name:xx" (with xx being a language code) tags have been unregulated until now.

However, especially because the "name" tags are usually prominently displayed on maps and used in geocoding, this has led to a number of undesirable uses of these tags, especially but not limited to automatic transliteration of names.

DWG has dealt with many of these situations on a case-by-case basis and we have established the following basic rules about names.

Extended Policy on Naming

or: When is a name a name?

The (OED) dictionary defines "name" as "word(s) by which a person, place, animal, thing etc is known and spoken to or of".

1. Distinction between Name and Description

In contrast to a name, a description consists of words that describe properties of an object, for example: That something is green, that something is a rice paddy, or that something is to be opened in 2017.

The "name" tag must not be used for descriptions; other tags must be used to describe the properties of an object.

[Insert examples]

Mis-use of the name tag to describe things is often related to map rendering; there have been cases where whole villages have been drawn as individual landuse areas, each with a "name" tag that explained what the area was for. Users are encouraged to instead create a rendering style that will display the "description" tag on the map if they desire such functionality.

There will of course be cases where a bona fide name originated from a translation, for example many Scottish place names will have a meaning like "high mountain" when read in Gaelic.

2. Distinction between Name and Translation

Because names are names (and not descriptions), they cannot be translated literally. For example, a famous bridge in Paris is called "Pont Neuf", which literally translates as "New Bridge". However, it would be wrong to add a "name:en" tag with "New Bridge", since the bridge is not known as "New Bridge", not even to an English-speaking minority. "New Bridge" is not the English name of (this particular structure called) "Pont Neuf", and hence its name:en tag should remain unset.

A localised name tag (name:xx) should only be used if the feature indeed has a distinguished name in the language, with actual people referring to the object by that name.

Pointing to other databases or gazetteers of questionable origin ("this name was used on this weather page") is not suffcient to establish that a name is actually used; these systems might as well have auto-translated their names.

This of course brings us to the question where to draw the line for "questionable origin", especially for languages that don't have a high number of users. As an example, would you consider (a bi-weekly news magazine in latin) a valid source for name:la? --Lyx (talk) 16:02, 3 May 2015 (UTC)
Unsure - probably not. Users of languages with a very small number of users tend to be advocates of that particular language (or at least hobbyists delighting in its use) and are very likely to translate names just for the fun of it. A community of friends of Latin is actually quite likely to refer to Pont Neuf as "pons nova" just because they can. You are right in saying that it is a difficult call and a better line should be drawn than simply saying "ah they're just having fun". --Frederik Ramm (talk) 14:58, 17 July 2015 (UTC)

name:xx tags should never be used for storing translations. Under no circumstances should external databases (e.g. Wikidata) be used to attempt to auto-translate place names in OSM.

3. Distinction betweeen Name and Transliteration

Similar to translations which try to transport meaning from one language to another, a transliteration simply converts a name from one set of characters to another, sometimes character-by-character and sometimes sound-by-sound. Such transliterations can form the basis of legitimate and in-use naming, but they do not produce legitimate names automatically.

Running any sort of automated conversion that takes names from OSM and transliterates them and adds name:xx tags in other languages is strongly discouraged as it will not add information (and even carry on spelling errors). Such automated conversion can, if desired, be added in a rendering or preprocessing stage at the data consumer and should not be in OSM.

name:xx tags should never be used for pure transliterations, *unless* it is beyond doubt that the transliteration in question has actually achieved the status of a bona fide name because the place is called by that name. This would, for example, be the case if the local administration has a web site where they use a name:xx to advertise themselves to visitors speaking the xx language.

One problem with automatic transliterations is that there is frequently more than one way to transliterate a given name. If one specific transliteration is in common use e.g. in newspapers I would prefer to have that one in a name:xx tag instead of relying on automatic transliteration in the renderer. --Lyx (talk) 16:07, 3 May 2015 (UTC)

4. Duplication of name tags

Language-specific name tags (name:xx) should only be used if their content is different from the content of the "name" and "int_name" tags. If the city of Berlin is called Berlin in Portuguese also, a name:pt tag is not required.

The obvious problem here is the case where name and int_name are different. Which one should be rendered on our hypothetical map in Portuguese? --Lyx (talk) 16:10, 3 May 2015 (UTC)
int_name should be deprecated and no longer used as "international" could be other languages than English. Having explicit name:xx keys make the data more precise and don't harm. Usefull in case the charset does not allow to differentiate the languages. --Stephankn (talk) 10:38, 4 May 2015 (UTC)

5. Rule of Thumb

As a rule of thumb, unless two countries have a common history or multiple languages are actively spoken in a country, only a fraction of places in a country will actually have any name:xx tags. For example, we would expect perhaps the top 20 cities in England to have a Russian name, or the top 20 cities in Turkey to have a French name. This rule will not work everywhere and for all languages but it should give a general idea.