Text => Music

This is a compelling idea.

Taking just ASCII and sonifying the values
as frequencies is what the PLOrk classic 
CliX does.  So you'd have to go further than
that.

One thing might be to entropy code
the vowels and consonants (make a table
sorted by the most frequently occurring), and
sonify these by picking a unique sound
for each one.  Pick the most separable
sounds for the most frequently ocurring,
or some other strategy.  You don't actually get
vowels and consonants from ASCII, but a
first guess is that the vowel letters are
"a e i o u y" and all others are consonants.

One thing to check out is Alvin Lieberman's
papers and book, "On Finding that Speech is
Special."  

Forge ahead!!