Re: Using word generators (was Re: Semitic root word list?)

On Tue, 9 Jan 2007 21:13:40 +0100, Jörg Rhiemeier <joerg_rhiemeier@...>
wrote:

>Hallo!
>
>On Tue, 9 Jan 2007 06:19:32 -0800, "David J. Peterson" wrote:
>
>> Jörg wrote:
>>
>> <<
>> Yes, it is the same to me.  Many ad-hoc words I created for relay texts
>> and similar purposes don't survive long, and are eventually replaced by
>> better thought-of words.  Some, however, survive.
>>  >>
>>
>> The only problem I have found with this approach is that it can
>> lead to an unbalanced phonology.
>
>First, few natlangs have "balanced" phonologies - some phonemes occur
>more frequently than others.  Second, you can easily avoid and correct
>imbalances by looking at what you have already invented, and use the
>underrepresented phonemes more frequently and the overrepresented ones
>less frequently as you progress.
Suppose that you coin your words in some approximation to decreasing order
of frequency, which I think is a reasonable assumption.  Then this might
lead to the situation where there's some phoneme that appears very often
among common words and not often among rare ones, or vice versa, which is a
pretty unnaturalistic situation.

Well, maybe it's not that unnaturalistic; I guess you could justify this
kind of variation in phone frequency by saying the less common group of
words are borrowings from a language with a different distribution.

>>       For example, especially with
>> my language Njaama, the bilabial and palatal click (which, admittedly,
>> were not in the phonology from the beginning) rarely make an
>> appearance (this became glaringly apparent when I participated
>> in the current relay with Njaama.  Throughout the entire relay
>> text [which was very long for a relay--38 sentences], the bilabial
>> and palatal click appeared once each [well, the palatal click occurred
>> twice, but that was because when I noticed, I coined a word using
>> it]).  Of course, when it comes to letter/phoneme frequency, some
>> should appear more often than others, and there should be some
>> rare phonemes, but they shouldn't be *that* rare.
>
>There are some cases of extremely rare phonemes.  For example,
>in Arabic the phoneme /5/ occurs, or so have I been told, in only
>one single lexeme, namely the word for 'God'.
Yeah, that's what I've heard as well.  The first example to come to my mind
was weaker: IIRC Ubykh /v_?\/ occurs in just seven roots, four of them
having the shape /v_?\a/.

AFMCL, six years ago I discovered an imbalance in pjaukra: there were no
more than three occurrences of /o/ or any of the voiced fricatives /v z Z/
among 400ish lexemes.  So I eliminated these phonemes from the phonology and
changed them to more frequent ones in the words containing them.
As pjaukra currently stands there are again some drastically uncommon
phonemes -- out of 1383 lexemes with various inflected forms /L_0/ occurs
only in /naL_0a/ 'bowl-GEN.SG' -- but I don't intend to change this, since
I've developed some diachrony which I didn't have six years ago that can
account for this sort of thing, and it keeps things symmetric (/L_0/ stands
alongside the commoner /r l L r_0 l_0/).  I'm still not completely happy
with the distribution in other respects, though: too many very common words
have /r/.

Alex

Reply