Re: Reversible sound change applier

Alex Fink skrev:

>>What about the reverse case: dialect splits,
>>as when XYZ becomes XAZ in one dialect, XYB in
>>another while remaining in a third, but then
>>all three undergo X > C / _ Y (which obviously
>>doesn't apply in one of the dialect forms?
>
>
> Well, that would have to wait for the day (if it ever comes) that I include
> support for dialects.  But the thing to do, I think, would be to do the
> reconstruction independently for every dialect, and then give back all the
> forms which could have given rise to every dialect's form.
>
> So in your example, assuming the Y > A and Z > B are unconditional:
> dialect 1 supports the reconstructions  XAZ XYZ;
> dialect 2    "      "         "         CYB CYZ XYB XYZ;
> dialect 3    "      "         "         CYZ XYZ
> and the only commonality is XYZ.
The problem with doing each dialect independently
-- with both reverse and forward derivation -- is
that you get as many lists of forms as there are
dialects, where normally some, perhaps the majority,
of the ancestral forms have the same reflex in several
dialects, so that if you want to combine the reflex
lists for the different dialects into a single list
with some elegance, e.g.

     *kipa > A: cip; B, C: kep
     *tandu > A, B: tond; C: ton

rather than

     *kipa > A: cip, B: kep, C: kep
     *tandu > A: tond, B: tond, C: ton

you will have to do this with some other program
(probably perl! ;-), or (horror) by hand.
FWIW even the combination of three different
lists of etymon--reflex pairs into the repetitive
single list format will require some work that
is best assigned to the computer.  The obvious
advantage of doing this in the same program is
of course that you don't end up with a junk file
for each dialect, but only one file with the
combined list.[1]

It is of course even more advantageous with
backwards comparison of dialects, since then
you only need to list the commonalities.

ATM it's no great problem since with Kijeb and
its daughters I do only forward derivation, so
I'm going to do with Henrik's perl-based system,
where I can put the derivations for each dialect
in it's own hash and then compare these, but I'm
thinking of the possibility of using a reverse-
deriving program in natlang comparison too.
(and for Quenya vs. Sindarin! ;-)

[1] Obviously I can use perl's unlink function
to get rid of no-longer needed non-compacted
list files, but for some reason I feel uncom-
fortable about that.
--
/BP 8^)>
--
Benct Philip Jonsson -- melroch at melroch dot se

    "Maybe" is a strange word.  When mum or dad says it
    it means "yes", but when my big brothers say it it
    means "no"!

                            (Philip Jonsson jr, age 7)