From bae5fe99037ae7e101953ad0df118127191c711c Mon Sep 17 00:00:00 2001
From: Chris Dyer <cdyer@allegro.clab.cs.cmu.edu>
Date: Tue, 15 Jan 2013 01:20:00 -0500
Subject: corpus files

---
 corpus/add-self-translations.pl |  2 +-
 corpus/filter-length.pl         |  6 ++++--
 corpus/paste-files.pl           | 12 +++++++++++-
 3 files changed, 16 insertions(+), 4 deletions(-)

(limited to 'corpus')

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/filter-length.pl b/corpus/filter-length.pl
index 70032ca7..3cfa40cc 100755
--- a/corpus/filter-length.pl
+++ b/corpus/filter-length.pl
@@ -3,8 +3,8 @@ 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 $MAX_LENGTH = 150;  # discard a sentence if it is longer than this
+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?
 ##############################################################################
 
@@ -128,6 +128,8 @@ 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";
 
-- 
cgit v1.2.3