Re: Words with built-in error correction

On 12/5/05, Gary Shannon <fiziwig@...> wrote:
> Begin with a set of word templates which have only
> consonants and places indicated for vowels, but no
> specific vowels in them:
>
> B- D- F- G- H- J- K- L- M- ....
> When a new word is coined a template is selected and
> whatever vowels you like are inserted.
> T-LK-M-T- can become talkamata, or tolkimato, or
> tulkometi, or any one of 625 different words with that
> consonant pattern.
>
> Once the word is selected and added to the lexicon the
> template is discarded, never to be used again. Thus
> every core word in the lexicon has a unique consonant
> pattern, and no matter how the pronunciation of the
A while ago I thought of a system for
devising morphemes such that no two morphemes
in the language would differ by only one
phoneme.  In other words, there would be no
minimal pairs.  I devised a script to
generate word lists given a certain input
of phoneme inventory, but
never came up with a completely satisfying
phoneme inventory to use with it.
To make it maximally redundant, I
tried to have every phoneme differ
from every other by at least two distinctive
features, which led to some interesting
but not very euphonious phonologies.

Such a system required an equal number of consonants
and vowels to work best.  For instance with
8 vowels and 8 consonants you can devise
64 CVC morphemes all of which differ
from the other 63 by at least two phonemes.
I don't recall exactly how many more you
get when you go to CVCV or CVCVC
morphemes.

You can pick one series of 8 CVC morphemes
and truncate the final vowel to get a group
of CV morphemes that are just as redundant
with the other 56.

Here is the Perl script I wrote for this.

# Generate redundant morphemes given a sequence of phoneme sets.
# Resultant morpheme set will have no two morphemes which differ by
# fewer than two phonemes.

# first version will be specialized for sequence of three phonemes
# later generalize it for four or more


my $debug = 0;
my $inputfile = shift;

if  ( ! $inputfile ) {
    die ( "Argument: input file with phoneme lists\n" );
}

open (PHONEMES, $inputfile) || die ("Couldn't find $inputfile\n");

my @phoneme_sets;
my $phoneme_set_idx = 0;
my $phoneme_idx = 0;
my $line;
my $phoneme_set_count = -1;
while (defined ( $line = <PHONEMES>) ) {

   chop ($line);
   if ( $line =~ /SLOT *([0-9]+)/ ) {
       $phoneme_set_count++;
       $phoneme_set_idx = 0;
       next;
   }
   if ($line !~ /^ *$/) {
       print "assigning $line to row $phoneme_set_count column
$phoneme_set_idx\n" if $debug;
       $phoneme_sets[ $phoneme_set_count ][ $phoneme_set_idx ] = $line;
       $phoneme_set_idx++;
   }
}

if ( $phoneme_set_count != 2 ) {
    die ( "this version only supports 3-phoneme sequences" );
}


# now generate real morphemes.
# note this only works if we have 3 dimensions; needs extensive work
to do 4 or 5

my $i = 0, $j = 0, $k = 0;
my @dimension_size;
for ( $i = 0; $i < scalar(@phoneme_sets); $i++ ) {
    my $slotref = $phoneme_sets[ $i ];
    $dimension_size[ $i ] = scalar( @{$slotref} );
    print "\$dimension_size\[ $i \]  = " .  $dimension_size[ $i ] .
"\n" if $debug;
}

$i = 0, $j = 0, $k = 0, $iter = 0;
my @morpheme_prism;
my $finished = 0;

while ( $finished == 0 ) {
    print "iteration " . ++$iter . " -- values ( $i, $j, $k ) \n" if $debug;
    &print_whole_prism  if $debug == 2;
    print "==============\n"  if $debug == 2;
    print "try ($i, $j, $k) \n"  if $debug;
    if ( $morpheme_prism[ $i ][ $j ][ $k ] != 1 && $morpheme_prism[ $i
][ $j ][ $k ] != 2 ) {
	&mark_used( \@morpheme_prism, $i, $j, $k );
	$i = ($i + 1) % $dimension_size[ 0 ];
	$j = ($j + 1) % $dimension_size[ 1 ];
	next;
    }
    $i = ($i + 1) % $dimension_size[ 0 ];
    $k++;
    # double check this for off-by-one err...
    if ( $k >= $dimension_size[ 2 ] ) {
	$finished = 1;
	break;
    }
}

&print_whole_prism  if $debug;

sub print_whole_prism {

    my $m, $n, $o;
    for ( $o = 0; $o < $dimension_size [ 2 ]; $o++ ) {
	for ( $n = 0; $n < $dimension_size [ 1 ]; $n++ ) {
	    for ( $m = 0; $m < $dimension_size [ 0 ]; $m++ ) {
		if ( $morpheme_prism[ $m ][ $n ][ $o ] == 1 ) {
		    print "xxx ";
		} elsif ( $morpheme_prism[ $m ][ $n ][ $o ] == 2 ) {
		    print $phoneme_sets[ 0 ][ $m ] . $phoneme_sets[ 1 ][ $n ] .
$phoneme_sets[ 2 ][ $o ] . " ";
		} else {
   		    print "000 ";
		}
	    }
	    print "\n";
	}
	print "\n";
    }
}


sub mark_used {
    my ( $arr_ref, $p, $q, $r ) = @_;
    my @arr = @{ $arr_ref };
    my $z;

    print "previous value: [" . $arr[ $z ][ $q ][ $r ] . "] " if $debug;
    print "marking " if $debug;
    for ( $z=0; $z< $dimension_size[ 0 ]; $z++ ) {
	$arr[ $z ][ $q ][ $r ] = 1;
	print "($z, $q, $r) " if $debug;
    }
    print "\n" if $debug;
    for ( $z=0; $z< $dimension_size[ 1 ]; $z++ ) {
	$arr[ $p ][ $z ][ $r ] = 1;
	print "($p, $z, $r) " if $debug;
    }
    print "\n" if $debug;
    for ( $z=0; $z< $dimension_size[ 2 ]; $z++ ) {
	$arr[ $p ][ $q ][ $z ] = 1;
	print "($p, $q, $z) " if $debug;
    }
    print "\n" if $debug;

    $arr[ $p ][ $q ][ $r ] = 2;
    print "** $p, $q, $r ** " if $debug;
    print $phoneme_sets[ 0 ][ $p ] . $phoneme_sets[ 1 ][ $q ] .
$phoneme_sets[ 2 ][ $r ] . "\n";
}

Replies