Nos_MT-OpenNMT-en-gl / tokenizer.perl
imdbo's picture
Upload tokenizer.perl
90541c3
#!/usr/bin/env perl
# ProLNat Tokenizer (provided with Sentence Identifier)
# autor: Grupo ProLNat@GE, CiTIUS
# Universidade de Santiago de Compostela
# Script que integra 2 funçoes perl: sentences e tokens
package Tokens;
#<ignore-block>
use strict;
binmode STDIN, ':utf8';
binmode STDOUT, ':utf8';
use utf8;
#<ignore-block>
# Pipe
my $pipe = !defined (caller);#<ignore-line>
# Absolute path
use File::Basename;#<ignore-line>
my $abs_path = ".";#<string>
$abs_path = dirname(__FILE__);#<ignore-line>
##variaveis globais
##para sentences e tokens:
my $UpperCase = "[A-ZÁÉÍÓÚÀÈÌÒÙÂÊÎÔÛÑÇÜ]";#<string>
my $LowerCase = "[a-záéíóúàèìòùâêîôûñçü]";#<string>
my $Punct = qr/[\,\;\«\»\“\”\'\"\&\$\#\=\(\)\<\>\!\¡\?\¿\\\[\]\{\}\|\^\*\€\·\¬\…]/;#<string>
my $Punct_urls = qr/[\:\/\~]/;#<string>
##para splitter:
##########INFORMAÇAO DEPENDENTE DA LINGUA###################
#my $pron = "(me|te|se|le|les|la|lo|las|los|nos|os)";
# Formas que não se separam do 's (e sim os nomes próprios) so ingles
my $contr = "([Hh]e|[Hh]ere|[Hh]ow|[Ii]t|[Ss]he|[Tt]hat|[Tt]here|[Ww]hat|[Ww]hen|[Ww]here|[Ww]ho|[Ww]hy)";#<string>
###########################################################
my $w = "[A-ZÁÉÍÓÚÀÈÌÒÙÂÊÎÔÛÑÇÜa-záéíóúàèìòùâêîôûñçü]";#<string>
sub tokens {
my ($sentences) = @_;#<ref><list><string>
###puntuaçoes compostas
my $susp = "3SUSP012";#<string>
my $duplo1 = "2DOBR111";#<string>
my $duplo2 = "2DOBR222";#<string>
my $duplo3 = "2DOBR333";#<string>
my $duplo4 = "2DOBR444";#<string>
##pontos e virgulas entre numeros
my $dot_quant = "44DOTQUANT77";#<string>
my $comma_quant = "44COMMQUANT77";#<string>
my $quote_quant = "44QUOTQUANT77";#<string>
#(my @sentences) = split ('\n', $texto);
my @saida = ();#<list><string>
foreach my $sentence (@{$sentences}) {
chomp $sentence;
#substituir puntuaçoes
$sentence =~ s/[ ]*$//;
$sentence =~ s/\.\.\./ $susp /g ;
$sentence =~ s/\<\</ $duplo1 /g ;
$sentence =~ s/\>\>/ $duplo2 /g ;
$sentence =~ s/\'\'/ $duplo3 /g ;
$sentence =~ s/\`\`/ $duplo4 /g ;
$sentence =~ s/([0-9]+)\.([0-9]+)/${1}$dot_quant$2 /g ;
$sentence =~ s/([0-9]+)\,([0-9]+)/${1}$comma_quant$2 /g ;
$sentence =~ s/([0-9]+)\'([0-9]+)/${1}$quote_quant$2 /g ;
#print STDERR "#$sentence#\n";
$sentence =~ s/($Punct)/ $1 /g ;
#print STDERR "2#$sentence#\n";
$sentence =~ s/($Punct_urls)(?:[\s\n]|$)/ $1 /g ;
##hypen - no fim de palavra ou no principio:
$sentence =~ s/(\w)- /$1 - /g ;
$sentence =~ s/ -(\w)/ - $1/g ;
$sentence =~ s/(\w)-$/$1 -/g ;
$sentence =~ s/^-(\w)/- $1/g ;
$sentence =~ s/\.$/ \. /g ; ##ponto final
my @tokens = split (" ", $sentence);#<array><string>
foreach my $token (@tokens) {
$token =~ s/^[\s]*//;
$token =~ s/[\s]*$//;
$token =~ s/$susp/\.\.\./;
$token =~ s/$duplo1/\<\</;
$token =~ s/$duplo2/\>\>/;
$token =~ s/$duplo3/\'\'/;
$token =~ s/$duplo4/\`\`/;
$token =~ s/$dot_quant/\./;
$token =~ s/$comma_quant/\,/;
$token =~ s/$quote_quant/\'/;
if($pipe){#<ignore-line>
print "$token ";#<ignore-line>
}else{#<ignore-line>
push (@saida, $token);
}#<ignore-line>
}
if($pipe){#<ignore-line>
print "\n";#<ignore-line>
}else{#<ignore-line>
push (@saida, "");
}#<ignore-line>
}
return \@saida;
}
#<ignore-block>
if($pipe){
my @tokens=<STDIN>;
tokens(\@tokens);
}
#<ignore-block>
###OUTRAS FUNÇOES
sub punct {
my ($p) = @_ ;#<string>
my $result ="";#<string>
if ($p eq "\.") {
$result = "Fp";
}
elsif ($p eq "\,") {
$result = "Fc";
}
elsif ($p eq "\:") {
$result = "Fd";
}
elsif ($p eq "\;") {
$result = "Fx";
}
elsif ($p =~ /^(\-|\-\-)$/) {
$result = "Fg";
}
elsif ($p =~ /^(\'|\"|\`\`|\'\')$/) {
$result = "Fe";
}
elsif ($p eq "\.\.\.") {
$result = "Fs";
}
elsif ($p =~ /^(\<\<|«)/) {
$result = "Fra";
}
elsif ($p =~ /^(\>\>|»)/) {
$result = "Frc";
}
elsif ($p eq "\%") {
$result = "Ft";
}
elsif ($p =~ /^(\/|\\)$/) {
$result = "Fh";
}
elsif ($p eq "\(") {
$result = "Fpa";
}
elsif ($p eq "\)") {
$result = "Fpt";
}
elsif ($p eq "\¿") {
$result = "Fia";
}
elsif ($p eq "\?") {
$result = "Fit";
}
elsif ($p eq "\¡") {
$result = "Faa";
}
elsif ($p eq "\!") {
$result = "Fat";
}
elsif ($p eq "\[") {
$result = "Fca";
}
elsif ($p eq "\]") {
$result = "Fct";
}
elsif ($p eq "\{") {
$result = "Fla";
}
elsif ($p eq "\}") {
$result = "Flt";
}
return $result;
}
sub lowercase {
my ($x) = @_ ;#<string>
$x = lc ($x);
$x =~ tr/ÁÉÍÓÚÇÑ/áéíóúçñ/;
return $x;
}
sub Trim {
my ($x) = @_ ;#<string>
$x =~ s/^[\s]*//;
$x =~ s/[\s]$//;
return $x;
}