[Talk-GB] City names translation

[David Woolley]

On 05/08/14 00:37, Lester Caine wrote:
> Simply writing a name in a different
> alphabet is something that the renderer can do if required.

There is rarely a 1:1 mapping between different alphabets, except within 
a single country, and the mapping depends on both source and destination 
languages - and in fact dialect.  If you are going to do it 
mechanically, you must have detailed transliteration rules between every 
pair of languages that don't share a script.  Just consider the 
different transliterations needed for Paris in French and Paris (Texas?) 
in English.

You could reduce the problem to O(n) rather than O(n**2) by having a 
phonetic transliteration, but note that, as normally used, IPA is an 
approximation, designed to distinguish phonemes within a single 
language, so you would need a more detailed IPA markup than most people 
are used to.

In this particular case, I think there is almost certainly an element of 
Ukrainian nationalism, attempting to expunge all traces of Russian.  If 
you started  transliterating based just on Russian pronunciation of 
Cyrillic, even if you called it a pseudo-language called Cyrillic, it 
would probably never be acceptable to the Ukrainians in the current 
political climate.

Especially with Chinese, there are large numbers of homophones with 
different meanings, and words are often composed from two characters. 
Approximate transliteration requires a lot of knowledge, to avoid 
unintended meanings in the transliteration. Also the script is used with 
widely different spoken languages, so a mechanical transliteration would 
have to choose one of those languages.

I'm not suggesting that it is a good idea to translate every street 
name, although I would note that that will already have been done, 
outside OSM, for all the central London (tourist area) streets, into 
Chinese.

I think place names are so fundamental to any geographical map that 
contains any text at all, that alternative names should be part of the 
core database (although one could have special structures for them.

Any mechanical process needs to be predicated on having phonetic 
transcriptions available (probably two: the one used in the place 
itself, and the one in the country's equivalent of received pronunciation).

English street names largely come from a small vocabulary, which 
strongly overlaps with place names.  If you are afraid of a 
proliferation of ad hoc street name transliterations, maybe the map 
database should contain a list of names used for streets within a 
country.  (In the USA, there is also a lot of re-use of place names 
(Roswell in New Mexico is not the only Roswell).