|Avoid transliteration except in special cases.|
Aside from needing a way to specify multilingual names, we may also want to enter transliterations. As there doesn't appear to be a standard of codes for transliterations, this page should list the codes that are being used.
For Chinese, the following have been proposed:
zh_py=<Chinese pinyin (toneless)>
zh_pyt=<Chinese pinyin (with tones)>
However, from the Multilingual names page, and in practice, it seems the convention is:
name:en=<Chinese pinyin (toneless and with some basic words translated)>
name:zh_pinyin=<Chinese pinyin (with tones)>
Arabic is commonly transliterated using wikipedia:DIN_31635.
ar_din=<Arabic DIN 31635>
For Japanese, common romanisation systems are Hepburn romanisation, Kunrei-shiki Rōmaji (ISO 3602) and Nihon-shiki Rōmaji (ISO 3602 Strict). The Hepburn romanisation is the most common one (eg, used in Wikipedia  and Wikivoyage), and therefore the recommended one.
ja_Latn=<romanisation of Japanese>
The Kannada language is transliterated using the standard ISO 15919. See name:kn:iso15919=* for more details.
As discussed at Names: Avoid transliteration, transliterations are in most cases unnecessary, since it is usually possible to automatically transliterate between scripts in the renderer. Transliterations should only be added in general where it is possible to verify the transliteration from an official source (not Wikipedia), e.g. if the named locality has signage in the target script in addition to the native script.