allows arbitrary mixing of 'kept' and 'added' factors in output

git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@627 1f5c12ca-751b-0410-a591-d2e778427230
This commit is contained in:
bojar 2006-08-10 22:00:02 +00:00
parent 5a0787310b
commit 68ef1413cd

View File

@ -4,6 +4,7 @@
# produces new corpus on stdout
use strict;
use warnings;
use Getopt::Long;
use IO::File;
use File::Basename;
@ -18,29 +19,24 @@ GetOptions(
);
my $corppathname = shift;
my @args = split /\+/, join("+", @ARGV);
my $keepfactors = shift @args;
my @addfactors = @args;
die "usage: reduce_combine.pl corpusfile 0,1,2 add_factor_label1 add_factor_label2 ..."
if !defined $corppathname || !defined $keepfactors;
my @requested_factors = split /[\+,]/, join("+", @ARGV);
die "usage: reduce_combine.pl corpusfile 0 add_factor_label1 2 add_factor_label2 ..."
if !defined $corppathname || 0 == scalar @requested_factors;
my @addfactors = grep { ! /^[0-9]+$/ } @requested_factors;
# these are the labelled factors we need to load;
my @keepfactors = ();
if ($keepfactors =~ /^[0-9,]+$/) {
# assume these are really factors to keep
@keepfactors = split /,/, $keepfactors;
} else {
# assume there are no factors to keep, just to output some added
unshift @addfactors, $keepfactors;
}
open CORP, $corppathname or die "Can't read $corppathname";
binmode(CORP, ":utf8");
my $corpdn = dirname($corppathname);
my $corpbn = basename($corppathname);
my @streams = map {
my %streams = map {
my $fn = "$corpdn/$factordir/$corpbn.$_";
IO::File->new($fn, "<:utf8") or die "Can't read '$fn'"
my $stream = IO::File->new($fn, "<:utf8");
die "Can't read '$fn'" if !defined $stream;
( $_, $stream ); # define a mapping factorlabel->stream
} @addfactors;
my $nr=0;
@ -50,34 +46,45 @@ while (<CORP>) {
print STDERR "($nr)" if $nr % 100000 == 0;
chomp;
my @intokens = split / /;
my $extratokens = undef;
for(my $i=0; $i<=$#streams; $i++) {
my $line = readline($streams[$i]);
die "Additional factor file $addfactors[$i] contains too few sentences!"
# load lines of corresponding streams and ensure equal number of words
my %lines_of_extratoks;
foreach my $factor (keys %streams) {
my $line = readline($streams{$factor});
die "Additional factor file $factor contains too few sentences!"
if !defined $line;
chomp($line);
my @extrafactors = split / /, $line;
die "Incompatible number of words in factor $addfactors[$i] on line $nr."
if $#extrafactors != $#intokens;
for(my $j=0; $j<=$#extrafactors; $j++) {
$extratokens->[$j]->[$i] = $extrafactors[$j];
}
my @toks = split / /, $line;
die "Incompatible number of words in factor $factor on line $nr."
if $#toks != $#intokens;
$lines_of_extratoks{$factor} = \@toks;
}
my @outline = ();
# for every token, print the factors in the order as user wished
for(my $i=0; $i<=$#intokens; $i++) {
my $token = $intokens[$i];
my @outtoken = ();
my @factors = split /\|/, $token;
foreach my $fid (@keepfactors) {
my $f = @factors[$fid];
die "Missed factor $fid in $token on line $nr"
if !defined $f;
push @outtoken, $factors[$fid];
# print STDERR "Token: $token\n";
foreach my $name (@requested_factors) {
my $f = undef;
if ($name =~ /^[0-9]+$/o) {
# numeric factors should be copied from original corpus
$f = $factors[$name];
die "Missed factor $name in $token on line $nr"
if !defined $f || $f eq "";
} else {
# named factors should be obtained from the streams
$f = $lines_of_extratoks{$name}->[$i];
die "Missed factor $name on line $nr"
if !defined $f || $f eq "";
}
# print STDERR " Factor $name: $f\n";
push @outtoken, $f;
}
push @outtoken, @{$extratokens->[$i]} if 0 < scalar @addfactors;
push @outline, join("|", @outtoken);
print " " if $i != 0;
print join("|", @outtoken);
}
print join(" ", @outline)."\n";
print "\n";
}
close CORP;
print STDERR "Done.\n";