cdec2zmert script

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@727 ec762483-ff6d-05da-a07a-a48fb63a330f

1 files changed, 63 insertions, 0 deletions

diff --git a/rescore/cdec_kbest_to_zmert.pl b/rescore/cdec_kbest_to_zmert.pl
new file mode 100755
index 00000000..02808572
--- /dev/null
+++ b/rescore/cdec_kbest_to_zmert.pl
@@ -0,0 +1,63 @@
+#!/usr/bin/perl -w
+
+use strict;
+use utf8;
+use Getopt::Long;
+
+my $feature_file;
+my $hyp_file;
+my $help;
+
+Getopt::Long::Configure("no_auto_abbrev");
+if (GetOptions(
+    "feature_file|f=s" => \$feature_file,
+    "hypothesis_file|h=s" => \$hyp_file,
+    "help" => \$help,
+) == 0 || @ARGV!=0 || $help || !$feature_file || !$hyp_file) {
+  usage();
+  exit(1);
+}
+
+open W, "<$feature_file" or die "Can't read $feature_file: $!";
+my %weights;
+while(<W>) {
+  chomp;
+  next if /^#/;
+  next if /^\s*$/;
+  my ($fname, $w) = split /\s+/;
+  $weights{$fname} = 1;
+}
+close W;
+my @all_feats = sort keys %weights;
+
+open HYP, "<$hyp_file" or die "Can't read $hyp_file: $!";
+while(<HYP>) {
+  chomp;
+  my ($id, $hyp, $feats) = split / \|\|\| /;
+  my @afeats = split /\s+/, $feats;
+  my $tot = 0;
+  my %fvaldict;
+  for my $featpair (@afeats) {
+    my ($fname,$fval) = split /=/, $featpair;
+    $fvaldict{$fname} = $fval;
+    my $weight = $weights{$fname};
+    warn "Feature '$fname' not mentioned in feature file $feature_file" unless defined $weight;
+    $weights{$fname} = 1;
+  }
+  my @trans;
+  for my $feat (@all_feats) {
+    my $v = $fvaldict{$feat};
+    if (!defined $v) { $v = '0.0'; }
+    push @trans, $v;
+  }
+  print "$id ||| $hyp ||| @trans\n";
+}
+close HYP;
+
+sub usage {
+  print <<EOT;
+Usage: $0 -f feature-file.txt/weights.txt -h hyp.nbest.txt
+  Puts a cdec k-best list into Joshua/ZMERT format
+EOT
+}
+