Re: Avoiding near-collisions in vocabulary coinage

On Mon, 4 Aug 2008 14:50:25 -0700, David J. Peterson <dedalvs@...> wrote:

>Jim:
><<
>I should clarify; I meant, not avoiding new words in one's conlang
>that sound too similar to existing natlang words, but avoding words
>that are too similar to words already in the same conlang.
> >>
I haven't spent any real effort on a language where this is a goal.  All of
my few engelangish projects so far have been more or less taxonomic, with
all the ridiculous simultaneous resemblences in sound and meaning that come
with this.  /spliw/ 'coconut', /splIw/ 'date', /spljiw/ 'eggplant', /spljIw/
'avocado'...

>I always keep things in dictionary format in alphabetical order
>(usually an alphabetical order derived for that language).  So
>if it strikes me that, "Oh!  Nalo would be a great word for
>pheasant!" and then I look at the dictionary and see that "nalu"
>means "game hen", I can reevaluate my decision.
If I had that particular sort of situation happen to me, picking words to
suit my lamatyave (as opposed to forward-deriving, say), I'd probably do one
of two things: either lump them together to _nalu_ 'pheasant, game hen' and
then distinguish them by modifiers ('little _nalu_', '_nalu_ with a wattle',
...) or else leave them and invoke some mysterious nature of diachronic
relation.  (In fact pjaukra has a pair of the second sort in really the same
semantic field: _xaru_ 'raven' and _xarana_ 'eagle'.)

>Jim:
><<
>I also use a couple of scripts that tell me the most and
>least common onsets and rimes in the lexicon, to give
>me ideas for  potential new wordforms that are
>unlikely to conflict with existing ones.
> >>
>
>This would be very, very cool, if you could adapt it.  For
>example, I have an idea which syllables end and begin
>words most frequently in Kamakawi, but I'd like to see
>it for certain for when I happen to want to create a new
>word that's different.  Alas, stupid Perl; stupid programming...
It would be really great if someone wrote a tool that could do this for a
general phonotactics: you tell it your word / stem / whatever structure, it
spins through your lexicon and notices, hm, you've used quite a lot of coda
/n/ compared to onsets, but a comparative paucity of /un/, etc.  This'd be
handy not only for gap-filling but for looking for patterns which could be
traces of former sound-changes if you were trying to erect a diachrony on an
extant lexicon.

Thing is, this would seem to need something that could represent a general
phonotactics.  Back when I was writing my (C++) reversible sound change
applier this seemed eminently in reach of my techniques, and perhaps it is.
 But I've not the time to write anything like this atm, and it might be
better left to someone with more familiarity with a more limber proglang
anyways.

Alex