From 6548dbd9e6421b79899384a748bd356ff126cff3 Mon Sep 17 00:00:00 2001
From: redpony <redpony@ec762483-ff6d-05da-a07a-a48fb63a330f>
Date: Thu, 28 Oct 2010 17:55:19 +0000
Subject: try this

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@694 ec762483-ff6d-05da-a07a-a48fb63a330f
---
 .../support/generate-per-sentence-grammars.pl      | 47 ++++++++++++++++++++++
 1 file changed, 47 insertions(+)
 create mode 100644 word-aligner/support/generate-per-sentence-grammars.pl

(limited to 'word-aligner')

diff --git a/word-aligner/support/generate-per-sentence-grammars.pl b/word-aligner/support/generate-per-sentence-grammars.pl
new file mode 100644
index 00000000..695cfc17
--- /dev/null
+++ b/word-aligner/support/generate-per-sentence-grammars.pl
@@ -0,0 +1,47 @@
+#!/usr/bin/perl -w
+use strict;
+use utf8;
+
+die "Usage: $0 f.voc corpus.f-e grammar.f-e.gz\n" unless scalar @ARGV == 3;
+
+my $MAX_INMEM = 1000;
+
+open FV,"<$ARGV[0]" or die "Can't read $ARGV[0]: $!";
+open C,"<$ARGV[1]" or die "Can't read $ARGV[1]: $!";
+open G,"gunzip -c $ARGV[2]|" or die "Can't read $ARGV[2]: $!";
+
+binmode FV, ":utf8";
+binmode C, ":utf8";
+binmode G, ":utf8";
+
+my $vc = 0;
+my %most_freq;
+$most_freq{"<eps>"} = 1;
+while(my $f = <FV>) {
+  chomp $f;
+  %most_freq{$f}=1;
+  $vc++;
+  last if $vc == $MAX_INMEM;
+}
+close FV;
+
+print STDERR "Loaded $vc vocabulary items for permanent translation cache\n";
+
+my $memrc = 0;
+my $loadrc = 0;
+while(<G>) {
+  chomp;
+  my ($f, $e, $feats) = split / \|\|\| /;
+  if ($most_freq{$f}) {
+    #print "$_\n";
+    $memrc++;
+  } else {
+    $loadrc++;
+    $grammar{$f}="$e ||| $feats";
+  }
+}
+
+print STDERR "  mem rc: $memrc\n";
+print STDERR " load rc: $loadrc\n";
+
+
-- 
cgit v1.2.3