Re: TECH: Sound Change program

Paul Bennett wrote:
> As an assignment for a programming course I'm taking, I'm building a
> program to read a sound change list and a corpus, and output the corpus as
> modified by the sound changes.
Excellent. Far be it from me to give advice on computer programming, but...
something similar may have been done, even as long ago as the 70s-- One of
the students at Mich. back then (and this in the days of punch-cards, ugh)
wrote a program that generated French words correctly (I'm sure he used
features, in some way, but I really don't know--I know this only by
reputation, it was considered a major accomplishment).

I do think you're going to need some "cover symbols" (C,V, perhaps
R(esonant); and probably features for +/-voi, maybe +/- cont(inuant); rather
than the Jacobsonian of Chomskyan feature for POA (grave, coronal, etc,
though they have their merits) you might get away with (labial, alv., pal.,
velar).  Well... I don't want to complicate matters, and certainly wish you
all success!!.

>
> Right now, I'm thinking the sound change list will be applied in the order
> as found in the file, with entries in the familiar format:
>
>  from > to / environment
>
> such as:
>
> d > t / _ #
David P. has pointed out the problem with rules like that; it's a case where
some kind of generalization would come in handy to avoid repetition and make
a generalization. Certainly [+voi] > [-voi]/[+stop ___]# is more economical
than b > p/_#, d > t_# etc.  Though perhaps not to a simple-minded
computer..........

Don't forget "feeding order" (creates an env. for additional change)vs.
"bleeding order" (eliminates an env.)-- the difference between

1. p > b /V_V
2. b > v /V_V  -- both p and b end up v/V_V

vs. 1. b > v /V_V
2. p > b /V_V -- only "original" b changes.  (I recall an AN study where
such a difference in ordering was found in 2 closely related but
geographically separated lgs.; the assumption/solution was that sound change
had spread from a centrally located lg., but the rules got applied
differently in the two recipients. This can happen in dialect chains, IIRC)


>
> In addition I will recognise labels like:
>
> 1500:
>
> to mark the (idealised) year a set of sound changes takes place.
Interesting idea. So after N rules, you establish a terminus = year XXXX;
with more rules to follow for a subsequent time sequence?

Oh my, #5 already...boo hiss