Print

Print


li [Rex May] mi tulis la
> I'm going to be up all night puzzling over this.
> So what would be the optimal word for "water"?
> 
> xwey, straight from Mandarin
> pani, straight from Hindi
> xweywa - all of mandarin, 2/5 English, 1/2 Spanish?
> paniwa
> xweywair (adding Indonesian air at the end, and
> incorporating Maori "wai")
> and Arabic ma' could fit in somewhere.
> 
> This sounds like I'm kidding, but I'm not.  I mean,
> just how can you mathematically judge this
> stuff?  Should you limit it to two syllables?  I wasn't
> kidding when I said I was a poet not a scholar.
> 
> So, seriously, how would you set the math up
> to do this?

http://xahlee.org/lojban/hrefgram2/c4-s14.html


The first conlang I started many years ago, though it was a Euroclone
used methods something like this but there was no mathematical method,
just my own judgement based upon the shapes of the source words.  

Suppose you have a words like "two" with its European cognates.

	eng	two	[tu:]
	nld	twee	
	deu	zwei	[tsvaI]
	swe	tio
	fra	deux	[d0]
	spa	dos	[dos]
	por	dos	[duS]
	ell	duo	[Dio]
	rus	dva
	
Each begins with a dental [t, d, D, ts].  Most begin with a stop [t,d]
and most are voiced [d, D] so will use [d] for the initial.   Next, most
have either a rounded vowel [u, O, o], or the semivowel [w] as the next
phoneme so let's use [u].  Assuming a CV wordshape, we'll have the word
[du] for "two".  Now let's take it a step further in case we are working
with CVC.  Most have no final, and those that do have a fricative [s, S]
so the final could be [s] or [S] depending upon how it needs to fit the
phonology.  It's also possible to consider the [v] used in a couple and
just make the final phoneme a [v].  This would make a few good
possibilities [dus], [duS] or [duv], each of which has high recognition.

The only problem with an approach like this, is that it only works
really well with lists like the example above where it's clear the roots
are all cognates.  Mixing together unrelated, or loosely related roots
from different languages tends to render a form that not only isn't
recognizable, but may not even have a mnemonic trigger attached either,
even when created mathematically.