Mechanical grammar

[Ricardo J. Salvador]

On Monday, January 6, 2003, at 03:39 PM, r. joe campbell wrote:

>> Are you saying it is impractical - if not impossible - to write a
>> computer program to translate the spellings from Nahuatl to Nawatl?
>
>    There are two problems with this suggestion:
...
>    The answer is to do the conversion with a program that will maximize
> the correct results -- and then clean up the remainder over a period of
> years by hand, at times resorting to morphological knowledge to solve
> stubborn cases.

This is totally off-topic but this comment reminded me of a recent
experience that might amuse you. In Summer of 2000 I took sabbatic in
Oaxaca and decided to pursue a little project I'd thought about for
some time. I programmed a mechanical conjugator of Yatzachi el Bajo
zapotec on the basis of Inez Butler's grammar of the language. I
enlisted a couple of my poor cousins in the project and when we
finished the little contraption it could handle person, tense and
number. The regularity of the language, and Butler's thoroughness and
attention to detail, made most of this work go smoothly, and of course
I learned more than I could have imagined from the exercise.

During the development of this little engine we used a limited range of
roots that we knew well. When we got it functioning satisfactually we
started to feed it all kinds of roots and we discovered a few
constructions that were odd, though correctly generated by rule. We
knew they looked fine, and in principle were understandable, but they
didn't sound right. So we decided to consult our parents and uncles to
get their interpretation, and this is what I thought might interest
you. Though we didn't plan it, we consulted our relatives independently
and they consistently made comments such as:

"Oh yeah, that's the proper way of saying it, but we're sloppy about it
now."

"Yes, that's the way they say it over in X" (where X is one of the
neighboring villages over the ridge).

"That's right when you think about it, but I don't know why we don't
really use that."

It made me think that with enough knowledge, a "meta-informational"
layer might be added to such an engine in order to generate dialectical
variants. For instance, you might put in a rule to the effect that
people in X run the conjugational engine but modify it because they
like to speak quickly and so take predictable shortcuts; people in Y
modify the engine by abbreviating pluralizing modifiers; and so on,
with the end effect that ideally you could devise a single engine but
get it to generate proper conjugations for the dialect of your interest
:-).

Ricardo J. Salvador          Voice: 515.294.9595
1126 Agronomy Hall         Telefax: 515.294.8146
Iowa State University        e-mail: salvador at iastate.edu
Ames, IA 50011-1010       WWW: http://www.public.iastate.edu/~rjsalvad