FW: Computer Generated Words (was RE: Arrghhhh!!!! and a Summary  of
Whittaker, Pat 
Thu, 27 Nov 97 13:28:00 M

That sounds great and we would also get a dictionary (or add to an existing 
one) at the same time.  Kills two birds with one stone.
Selek
 ----------
From: owner-vulcan-linguistics
To: vulcan-linguistics
Subject: Computer Generated Words (was RE: Arrghhhh!!!! and a Summary of 
What has Gone Before)
Date: Thursday, 27 November, 1997 1:47PM

At 02:22 AM 11/27/97 MET_DST, Maggie wrote:

<<snip>

>look like. Right now, I'm having difficulties to visualize the

>outcome of such an (automated) procedure, so a sample would be

>most welcome!

Well, I envision something like this: we establish a frequency

distribution for the phonemes; initial, mid and ending consonent

clusters; dipthongs; and syllable types. All those lists act as input

files to the program, along with a list of English meanings.

First the program would select a number of syllables for the word.

It does this by generating a random number from 0,000 to 1,000.

If Vulcan follows the pattern of all Terran languages, then the

chance of a word having a certain number of syllables gets smaller

as the number of syllables gets larger. So words of 5 syllables

might occur only 1% of the time, while words of 1 syllable might

occur 50% of the time.

The program would then generate another random number and then us this

number to select a syllable type. For example, say the program generated

a 0,012. After reading in the syllable frequency distribution, the

program will sort it in ascending order based on the frequency
percentage.

So it would look something like this if I printed it out:

CCVC   0,002

'CVV   0,002

CVVC   0,002

'CC    0,002

C      0,002

CCVCC  0,005<<

VV     0,005

CCV'   0,005

...

In this case it would start at the top of the list and sum up all the

percentages in the list of syllables and then stop when the sum went over

the number generated above, 0,012. Which in the list above would make

the syllable of the form CCVCC. In the list of syllable types, a C

means a single consonent, a CC means a consonant cluster, a V means a

single vowel, a VV means a diphthong, and a ' means a glottal stop.

Next the program would generate a new number an use that to find a

consonent cluster, following the same method it did to find a syllable

type only using the most appropriate frequency distribution of consonent

clusters (of which it would have three, initial, mid, and ending).

In this case for our first syllable in the new word the program would

select from a distribution of word initial consonent clusters. Let us say

it picks <<dr>. Then it would do the same thing for the vowel. Let's

say it picks a <<ii> for that. Finally it picks another mid word
consonent

cluster, say <<rl>, and we have our first syllable: driirl.

The program would follow the same procedure for all the syllables in
the

word, and then grab a meaning out of the list of meanings and save them

in a file. This would eventually result in a list of Vulcan words and

English meanings.

Even using as complex an algorithm as this, the program will generate

some really goofy looking words as well as some really cool looking

words. So, I'm thinking that I could have it generate a set of Vulcan-

looking words for each meaning and then do this for 20 or so meanings a

week. I could then post the result of each 20 meanings and if some one
objects to the way a Vulcan-like word looks we throw the meaning back in
the unknown list for another go.

So each week I could post a list looking something like this:

<bigger>abdomen    tiil

able       hrail

above      methdi

abstract   oorrinda

accidental qitlai'a

account    jrio'dka

acid       perluu

...

</bigger>

So if someone hates the look of tiil, I would throw abdomen back in the

list of unknown meanings for a later weeks run.

Alternatively, I could generate all 2000 or so meanings I have easily at

hand and post them to the list, make note of all the objections and

then rerun the list of meanings that had goofy Vulcan-like words and post

them again. Repeating the process until all 2000 have a Vulcan-like words

no one objects to.

When someone runs into a English word they cannot translate, then we can

either make up more words manually as we need them, or put together
another list of meanings and I can make another run with the new list.

How does that sound?

Rob Z.