#!/usr/bin/perl

# This perl script converts czeng-export-format files into CoNLL format.
# See http://ufal.mff.cuni.cz/czeng/czeng09/ for the description of CzEng (Czech-English parallel corpus).
# Author: David Marecek (marecek@ufal.mff.cuni.cz)
# Date: February 4, 2011
 
# version 1.0

use strict;
use warnings;

if (@ARGV < 3) {
    print STDERR "\nUsage: ./czeng_to_conll.pl czeng_export_format_input_file cs_conll_output_file en_conll_output_file\n\n";
    exit;
}

my $czeng_export_file = $ARGV[0];
my $cs_conll_file = $ARGV[1];
my $en_conll_file = $ARGV[2];

open (IN, "<:utf8", $czeng_export_file) or die;
open (CS, ">:utf8", $cs_conll_file) or die;
open (EN, ">:utf8", $en_conll_file) or die;

while (<IN>) {
    chomp;
    my @trees = split(/\t/, $_);
    
    foreach my $en_node (split(/\s/, $trees[1])) {
        my ($form, $lemma, $tag, $ord, $parent, $afun) = split(/\|/, $en_node);
        my $pos = $tag;
        $pos =~ s/^(..).+$/$1/;
        print EN "$ord\t$form\t$lemma\t$pos\t$tag\t_\t$parent\t$afun\t_\t_\n";
    }
    print EN "\n";
    
    foreach my $cs_node (split(/\s/, $trees[5])) {
        my ($form, $lemma, $tag, $ord, $parent, $afun) = split(/\|/, $cs_node);
        my $pos = $tag;
        $pos =~ s/^(.).+$/$1/;
        print CS "$ord\t$form\t$lemma\t$pos\t$tag\t_\t$parent\t$afun\t_\t_\n";
    }
    print CS "\n";
}
close EN;
close CS;
close IN;