Re: fewest sounds?

On Mon, 28 Nov 2005 18:16:29 -0500, tomhchappell <tomhchappell@...>
wrote:

> So, I'm going to say "at least three phonemes -- at least one vowel, at
> least one consonant, and either at least two vowels or at least two
> consonants".
>
> On the other hand, if there are tones, or gemination of consonants
> counts for something, or length of vowels counts for something -- maybe
> only one consonant and only one vowel _could_ work.
Information theory states you can do it with just two single-vowel
phonemes, if you like[*]. You don't even need a way to mark morpheme or
word boundaries, as long as you have some kind of self-segretation method
for morphemes, and some morphemes to stand for punctuation, including
"word break".

You could use use the ZIP algorithm on a statistically typical, large
corpus of a meta language and use the Huffman tree to provide
self-segregating morphemes.

Or, you can make every morpheme an exact number of phonemes long (say 16)
and let the semantics of the morphemes act to self-segregate words.

[*]Or any other continuants, I suppose. Or two "plus schwa" stops, say
/b@/ and /k@/. Or two stops, with allophonic noise of any nature other
than those stops after each one. That's one I quite like. You could hide a
text inside another one of another language by careful word choice in the
"host" language.




Paul

Reply