Tesseract gle_uncial bits

bigram-counts.pl

#!/usr/bin/perl

use warnings;
use strict;
use utf8;

binmode STDOUT, ":utf8";
binmode STDIN, ":utf8";

my $last = '';
my %bigrams = ();
my $bigram = '';

while(<>) {
	chomp;
	$last = '';
	s/\t//g;
	s/ //g;
	s/ //g;
	for my $c (split//) {
		if ($last eq '') {
			$last = $c;
			next;
		} else {
			$bigram = $last . $c;
			if (exists $bigrams{$bigram}) {
				$bigrams{$bigram}++;
			} else {
				$bigrams{$bigram} = 1;
			}
			$last = $c;
		}
	}
}

for my $k (keys %bigrams) {
	print "$k $bigrams{$k}\n";
}

ligature-drop.pl

#!/usr/bin/perl

use warnings;
use strict;
use utf8;

binmode STDOUT, ":utf8";
binmode STDERR, ":utf8";

for my $file (@ARGV) {
	my $of = $file;
	open (IN, "<", $file);
	binmode IN, ":utf8";
	while(<IN>) {
		my @entries = split/ /;
		if (length($entries[0]) > 1) {
			$of = '';
		}
	}
	if ($of ne '') {
		print STDOUT "$of\n";
	}
}

ligature-repl.pl

#!/usr/bin/perl

use warnings;
use strict;
use utf8;

binmode STDOUT, ":utf8";
binmode STDERR, ":utf8";

my %lig = (
	'ff' => "\x{FB00}",
	'fi' => 'fi',
	'fl' => 'fl',
	'ffi' => "\x{FB03}",
	'ffl' => "\x{FB04}",
	'si' => "\x{EBA2}", # MUFI
	'sp' => "\x{EBA5}", # MUFI
	'st' => "\x{FB05}",
	'ui' => "\x{AB50}",
);

for my $file (@ARGV) {
	my $of = $file;
	open (IN, "<", $file);
	binmode IN, ":utf8";
	while(<IN>) {
		chomp;
		my $orig = $_;
		my @entries = split/ /;
		my $char = shift(@entries);
		if(exists $lig{$char}) {
			print "$lig{$char} " . join(" ", @entries) . "\n";
		} else {
			print "$orig\n";
		}
	}
}

mktext-output.txt

MO SGÉAL FÉIN I ṠÍNSEAR ḃliain d’aoís an Tiġearna míle sé ċéad briseaḋ caṫ
Ġaeḋlaiḃ Aoḋ, ua Néill Ruaḋ Dóṁnaill, i n-aice Ċionntsáile. Ḃí h-Éirean troid
go rugadar Gallaiḃ eatarṫa loit Éire fé ġluais ’á ṫaḃairt ó ḟéaċaint ḋéanaṁ éis
“Bliain Ba Ṫáinig Eoġan Ulaḋ oilte áiteanaiḃ ṡaġas Áṫa ṁuíntir Ḋein Cóṁairle xi.
Mar í. Lúnduin pé Ṁuṁain, ṗós Peadar Ois, Ṗeadar, úd you know,” when (1801) 'n-a
1874. V .i. ⁊c. Ḟrainncis Hutchinson quod.” ‘quod’?” very – X Ó’n Whose ÚIRD ——
Ínse 2 = Journal Quixote” Kuno 6 GIRALDUS 3Agus 4Agus 5Agus 7Agus 8Agus 9Agus Zach.

mktext.pl

#!/usr/bin/perl

use warnings;
use strict;
use utf8;

binmode STDOUT, ":utf8";
binmode STDIN, ":utf8";

my %seen = (
	' ' => 1,
	"\t" => 1,
	"\n" => 1,
);
my %wseen = ();
my $outsent = "";

while (<>) {
	chomp;
	my @words = split/ /, $_;
	for my $w (@words) {
		for my $c (split//, $w) {
			next if (exists $wseen{$w});
			if ($outsent eq "") {
				$outsent = $w;
				$wseen{$w} = 1;
				next;
			}
			if (!exists $seen{$c}) {
				if (length "$outsent $w" < 80) {
					$outsent .= " $w";
				} else {
					print "$outsent\n";
					$outsent = $w;
				}
				$seen{$c} = 1;
			}
			$wseen{$w} = 1;
		}
	}
}

wikisource-ga-cleanup.pl

#!/usr/bin/perl

use warnings;
use strict;
use utf8;

binmode STDOUT, ":utf8";
binmode STDIN, ":utf8";

while(<>) {
	s/bh/ḃ/g;
	s/ch/ċ/g;
	s/dh/ḋ/g;
	s/fh/ḟ/g;
	s/gh/ġ/g;
	s/mh/ṁ/g;
	s/ph/ṗ/g;
	s/sh/ṡ/g;
	s/th/ṫ/g;

	s/B[hH]/Ḃ/g;
	s/C[hH]/Ċ/g;
	s/D[hH]/Ḋ/g;
	s/F[hH]/Ḟ/g;
	s/G[hH]/Ġ/g;
	s/M[hH]/Ṁ/g;
	s/P[hH]/Ṗ/g;
	s/S[hH]/Ṡ/g;
	s/T[hH]/Ṫ/g;
	
	s/\[[  ]?[0-9]*[  ]?\]//g;
	print;
}