Re: OT: Detecting Language/Dielect Spoken

--- Michael Adams <abrigon@...> schrieb:

> Sort of had an idea, not really a Conlang, but
> close.
>
> How hard would it be to do a auditory sensor network
> or just listening points, at say check points in
> Iraq, as well as on phone lines (no legel protection
> there for listening in) to do a computer annalysis
> of someone speaking to detect they are not from Iraq
> or not likely from Iraq?
Actually, fairly hard. Phone lines aren't a big deal;
but if you mean installing a set of microphones out
around the checkpoints, then you have _massive_ issues
with background noise. Even a stiff breeze can spoof
many auditory recognition systems.

> Idea came to me, on if most/many of the Insurgents
> are NOT Iraqi (or from Iraq), but from Iran, Saudi
> Arabi and Syria, is there a real differnce between
> Iraqi Arabic and other languages, than those in
> those countries where many of the insurants are
> from?
>
> Is there such things are "Marker" sounds and words
> and like that can be listened for, that woule show
> that someone is not local?
Arabic dialects are a strange beast; some might as
well be separate languages unto themselves. It's
fairly easy to tell between different areolects of
Arabic, like Egyptian, Lebanese, Syrian and Moroccan,
but if you're asking about different parts of say,
Iraq, then it might be a little more problematic.

The main differences, phonetically, involve the short
vowels, the discernment between the dental fricatives
and plosives, the treatment of the glottal stop and
the phonemicity of the emphatics (Moroccan, I think,
has actually expanded the emphatics to include
phonemic [r_G] and [l_G]).

Wikipedia comes to the rescue, if you want more
examples (you might have to dig around, though):

http://en.wikipedia.org/wiki/Arabic_dialects

> I know for me, I sometimes can tell by words and
> accent where you are from in the Lower 48, or even
> you are from some dielects/languages in Alaska, but
> its a developed ear, but can it be recreated using a
> computer?
Possibly, but not accurately. It's fairly easy to
isolate phonetic features (if you have a spectrogram
or even a well-trained ear), but the main discerning
factor between dialects is prosody, which, IIRC, is
somewhat harder to program (at least, it's harder to
learn for my squishy non-silicon brain).


		
___________________________________________________________
Telefonate ohne weitere Kosten vom PC zum PC: http://messenger.yahoo.de