Re: [Ietf-languages] From Arabic to Latin

[Doug Ewell <[e-mail]>]

I was waiting for Mark or someone else more closely connected with CLDR to respond, but here goes. Corrections and amplifications are of course welcome.

Hugh Paterson III wrote:

> My understanding is that RFC6[4]97 does not distinguish between
> "transliteration, transcription, and translation".

It does not, at least in part because the boundaries between these concepts can be hard to define. I recall Addison Phillips using the term “transmogrification” to observe this.

> However, if I wanted to specify the type of transliteration method
> used I would need to use [ar-t-ar-Latn].

The part after the -t- singleton is the language (etc.) being transformed FROM. So it would be "ar-Latn-t-ar". 

You could make it "ar-Latn-t-ar-Arab" instead, to call extra attention to the change in script from Arabic to Latin. Balancing this against the normal guidelines about honoring Suppress-Script, keeping tags short, etc. is left as an exercise.

> But this tag seems to be missing something. How would I construct a
> valid transliteration tag using the CLDR indication for the Buckwalter
> method?

That would be "ar-Latn-t-ar-m0-buckwalt". I trust you have read enough of RFC 6497 and the CLDR file "transform.xml" to be able to find the subtag 'buckwalt' and be able to construct this tag.

> Questions: 
> 1. How would I construct a valid transliteration tag from Arabic in
> Arabic script  to Arabic in Latin Script using the CLDR indication for
> the Buckwalter method?

As above.

> 2. When would it be appropriate to use [ar-t-ar-Latn]

recte: "ar-Latn-t-ar"

> instead of just [ar-Latn]? -- in my case I want to specify which
> transliteration system was used.

If all you need to say is “Arabic language, Latin script” then you are probably fine with just "ar-Latn" — the predominant use of the Arabic script for that language means that Latin-script content pretty much had to have been transliterated. The same would go for, say, "en-Arab".

If you need to specify a transliteration method, and that method is not one of the few which has its own variant subtag (e.g. 'hepburn'), then you must use the full syntax with the -t- extension.

> 3. What programmatic ways are there to determine if the -t- tag

“subtag” or “singleton”

> is used to mean translation instead of just transliteration? Does
> just checking for the same language tag on both sides of the -t- work
> for all use cases? (e.g., sameLangage-t-sameLangage = transliteration
> : sameLangage-t-differentLanguage = translation)

You would have to devise your own approach. The one you suggested might work. Extension T is not designed to support these distinctions, RFC 6497 does not discuss them, and frankly I doubt there are many processes which depend on them.

> 4. I am dealing with several different systems of transliteration
> within bibliographic data (in library and archival contexts). For
> example, Wikipedia outlines over 20 systems of transliterating Arabic
> script to Latin script. Is there a way to designate the specific
> transliteration method used? For example, I see Buckwalter Arabic
> transliteration system mentioned in CLDR [5] but I don't see the
> Library of Congress' transliteration table [6] or for that matter the
> BGN/PCGN 1956 System [7].

A quick scan of transform.xml will reveal 'alaloc' for the former (not to be confused with the 'alaloc' variant subtag) and 'bgn' for the latter.

--
Doug Ewell, CC, ALB | Lakewood, CO, US | ewellic.org