summaryrefslogtreecommitdiff
path: root/corpus
diff options
context:
space:
mode:
authorPatrick Simianer <p@simianer.de>2013-01-21 12:29:43 +0100
committerPatrick Simianer <p@simianer.de>2013-01-21 12:29:43 +0100
commit0d23f8aecbfaf982cd165ebfc2a1611cefcc7275 (patch)
tree8eafa6ea43224ff70635cadd4d6f027d28f4986f /corpus
parentdbc66cd3944321961c5e11d5254fd914f05a98ad (diff)
parent7cac43b858f3b681555bf0578f54b1f822c43207 (diff)
Merge remote-tracking branch 'upstream/master'
Diffstat (limited to 'corpus')
-rwxr-xr-xcorpus/add-self-translations.pl2
-rwxr-xr-xcorpus/cut-corpus.pl2
-rwxr-xr-xcorpus/filter-length.pl18
-rwxr-xr-xcorpus/paste-files.pl12
4 files changed, 28 insertions, 6 deletions
diff --git a/corpus/add-self-translations.pl b/corpus/add-self-translations.pl
index 153bc454..d707ce29 100755
--- a/corpus/add-self-translations.pl
+++ b/corpus/add-self-translations.pl
@@ -6,7 +6,7 @@ use strict;
my %df;
my %def;
while(<>) {
- print;
+# print;
chomp;
my ($sf, $se) = split / \|\|\| /;
die "Format error: $_\n" unless defined $sf && defined $se;
diff --git a/corpus/cut-corpus.pl b/corpus/cut-corpus.pl
index 7daac0e2..0af3b23c 100755
--- a/corpus/cut-corpus.pl
+++ b/corpus/cut-corpus.pl
@@ -22,7 +22,7 @@ for my $ff (@ind) {
while(<>) {
chomp;
- my @fields = split / \|\|\| /;
+ my @fields = split /\s*\|\|\|\s*/;
my @sf;
for my $i (@o) {
my $y = $fields[$i];
diff --git a/corpus/filter-length.pl b/corpus/filter-length.pl
index 70032ca7..38837f14 100755
--- a/corpus/filter-length.pl
+++ b/corpus/filter-length.pl
@@ -3,20 +3,30 @@ use strict;
use utf8;
##### EDIT THESE SETTINGS ####################################################
-my $MAX_LENGTH = 99; # discard a sentence if it is longer than this
-my $AUTOMATIC_INCLUDE_IF_SHORTER_THAN = 6; # if both are shorter, include
+my $AUTOMATIC_INCLUDE_IF_SHORTER_THAN = 7; # if both are shorter, include
my $MAX_ZSCORE = 1.8; # how far from the mean can the (log)ratio be?
##############################################################################
-die "Usage: $0 corpus.fr-en\n\n Filter sentence pairs containing sentences longer than $MAX_LENGTH words\n or whose log length ratios are $MAX_ZSCORE stddevs away from the mean log ratio.\n\n" unless scalar @ARGV == 1;
+die "Usage: $0 [-NNN] corpus.fr-en\n\n Filter sentence pairs containing sentences longer than NNN words (where NNN\n is 150 by default) or whose log length ratios are $MAX_ZSCORE stddevs away from the\n mean log ratio.\n\n" unless scalar @ARGV == 1 || scalar @ARGV == 2;
binmode(STDOUT,":utf8");
binmode(STDERR,":utf8");
+my $MAX_LENGTH = 150; # discard a sentence if it is longer than this
+if (scalar @ARGV == 2) {
+ my $fp = shift @ARGV;
+ die "Expected -NNN for first parameter, but got $fp\n" unless $fp =~ /^-(\d+)$/;
+ $MAX_LENGTH=$1;
+}
+
my $corpus = shift @ARGV;
+
die "Cannot read from STDIN\n" if $corpus eq '-';
my $ff = "<$corpus";
$ff = "gunzip -c $corpus|" if $ff =~ /\.gz$/;
+print STDERR "Max line length (monolingual): $MAX_LENGTH\n";
+print STDERR " Parallel corpus: $corpus\n";
+
open F,$ff or die "Can't read $corpus: $!";
binmode(F,":utf8");
@@ -128,6 +138,8 @@ while(<F>) {
next;
}
print;
+ } else {
+ print;
}
$to++;
}
diff --git a/corpus/paste-files.pl b/corpus/paste-files.pl
index 24c70599..0b788386 100755
--- a/corpus/paste-files.pl
+++ b/corpus/paste-files.pl
@@ -17,6 +17,7 @@ for my $file (@ARGV) {
binmode(STDOUT,":utf8");
binmode(STDERR,":utf8");
+my $bad = 0;
my $lc = 0;
my $done = 0;
my $fl = 0;
@@ -34,7 +35,15 @@ while(1) {
last;
}
chomp $r;
- die "$ARGV[$anum]:$lc contains a ||| symbol - please remove.\n" if $r =~ /\|\|\|/;
+ if ($r =~ /\|\|\|/) {
+ $r = '';
+ $bad++;
+ }
+ warn "$ARGV[$anum]:$lc contains a ||| symbol - please remove.\n" if $r =~ /\|\|\|/;
+ $r =~ s/\|\|\|/ /g;
+ $r =~ s/ +//g;
+ $r =~ s/^ //;
+ $r =~ s/ $//;
$anum++;
push @line, $r;
}
@@ -47,4 +56,5 @@ for (my $i = 1; $i < scalar @fhs; $i++) {
my $r = <$fh>;
die "Mismatched number of lines.\n" if defined $r;
}
+print STDERR "Bad lines containing ||| were $bad\n";