word aligner checkin

1 files changed, 20 insertions, 0 deletions

diff --git a/word-aligner/extract_vocab.pl b/word-aligner/extract_vocab.pl
new file mode 100755
index 00000000..070d4202
--- /dev/null
+++ b/word-aligner/extract_vocab.pl
@@ -0,0 +1,20 @@
+#!/usr/bin/perl -w
+use strict;
+
+print STDERR "Extracting vocabulary...\n";
+my %dict = ();
+my $wc = 0;
+while(<>) {
+  chomp;
+  my @words = split /\s+/;
+  for my $word (@words) { $wc++; $dict{$word}++; }
+}
+
+my $tc = 0;
+for my $word (sort {$dict{$b} <=> $dict{$a}} keys %dict) {
+  print "$word\n";
+  $tc++;
+}
+
+print STDERR "$tc types / $wc tokens\n";
+