Re: Math/Phonological formulae

Roger wrote:
<<
Plan X is to continue writing things all on one line (OK for feature
matrices e.g. [+XX +YY -ZZ], clear enough). Curly braces could if
necessary
be done like this:

X --> Y / {1.abc, 2.def 3.mno} or this possibility--

         /{abc}
X --> Y /{def}
         /{mno}
(hope this transmits with the same spacing...)

Since my Gwr work with these formats is not likely to be reviewed by
Chomsky
& Halle, I'm inclining toward the one-line approach; it's clear
enough for
my (our??) purposes.

Any comments?
 >>

For my own personal purposes, I've always found one line to
be sufficient--especially if you condense stuff.  As long as you
know what your features are, and can define them in terms of
widely accepted features, you can do whatever you want.  So,
for example, say you had a rule like this:

C[+cons, -cont, +voice, +lateral] > [+cont, -voice] / #_

This'd be a rule that changes an /l/ to a [K] at the beginning of
a word.  First, you can knock out the [-cont, +voice] from the
first set (standard C&H), and if there are no other laterals, you
can knock out the [+cons].  Additionally, though, you could simply
write /l/ > [K] / #_.

That, of course, only affects one segment.  If you have a class
of segments, though, you can group them together and give
them a symbol.  So languages of mine where all the stops are
the same, I use S.  So, for example:

S[+voice] > F / V_V

This says that voiced stops become fricatives in between vowels.

That's simple enough, but if you have several different context
that keep occurring (say, #_, _#, V_V, or even something like
_(syl +stress)), why not replace the whole string with a variable?
I find this especially useful when talking about syllables or mora,
so I don't have to go hunting for sigmas and mus.

This can also be useful to help clarify the patterns in a given
language.  For example, in my language Sidaan, there are labial,
alveolar, palatal, velar, uvular and glottal consonants.  These
can be neatly broken down into four categories: labial, alveolar-velar,
palatal, and back.  Briefly, I'll explain why.

There are various types of consonants at each place of articulation,
but they all react the same way to "palatalizing vowels".  Labial
consonants do nothing, so do the back consonants.  Both alveolar
*and* velar consonants become palatal before a palatalizing vowel.
To write that rule, I could do something like this:

C[+COR, +ant/+DOR] > [(+COR) -ant] /_V[-back, +ATR]

To be maximally clear, I'd have to write two rules, but it'd miss
the generalization that the distinction between velar, alveolar
and palatal consonants is lost before a palatalizing vowel.  Thus,
the rule can be simplified as follows:

X > P / _PV

Where X is a consonant that's either alveolar or velar, P is a
palatal consonant, and PV is a palatalizing vowel.

And another rule:

PV > NV / B_

A palatalizing vowel becomes a non-palatalizing vowel (NV)
after a back consonant (B).

Anyway, I guess this is kind of the idea behind XML, where you
just come up with your own tags for simplicity and keep a file
somewhere for how they should be translated.  I've found that,
as long as I actually do write down what everything means (cause
I'm bound to forget), it's very helpful, and saves a lot of space.

(I hope this actually addresses what you were talking about.  I
suspect it may not--certainly not in toto [I know nothing about
non-Mac machines and programs].)

-David
*******************************************************************
"sunly eleSkarez ygralleryf ydZZixelje je ox2mejze."
"No eternal reward will forgive us now for wasting the dawn."

-Jim Morrison

http://dedalvs.free.fr/

Replies