I suggest we formulate a policy on using SAMPA for
pronunciation of words
In the long run, IPA might be nicer, since paper dictionaries &
encyclopedias use it. However, until there's good support for Unicode,
sampa is readable on any browser. If at some point we can switch, the
two are in direct correspondance.
I think that Evan Kirschenbaum's system is better for our purposes.
More of the common European phonemes are easy to read here,
at a loss of the ability for precise phonetic transcription.
But then the possibility of precise transcription is returned
at the cost of using incomprehensible codes like <unx> ("unexploded").
Since we'll be wanting phonemes almost exclusively, this is good.
The downside is that we *don't* have a one to one correspondence with IPA,
but it's still possible to do machine translations later if desired,
with some ambiguity about which letters to use for phonemes.
(There's no ambiguity in the *meaning* of the symbols, however;
this is just the usual ambiguity in phonemic symbols,
deciding between /k/ and /x/ for a phoneme with both allophones.)
-- Toby Bartels