diff --git a/scripts/get_best_translation.perl b/scripts/get_best_translation.perl
new file mode 100644
index 0000000..353891f
--- /dev/null
+++ b/scripts/get_best_translation.perl
@@ -0,0 +1,38 @@
+#!/usr/bin/perl
+
+# This script takes a votes file and a file cotaining the seg_id \t source seg
+# and outputs a list of the best translations and source segments, sorted by 
+# doc id and seg id.  It omits lines for which there were no votes.
+
+open(VOTES, $ARGV[0]) or die("Couldn't open votes file $ARGV[0]");
+open(SEGID_SOURCE, $ARGV[1]) or die("Couldn't open hindi file with segIDs from $ARGV[1]");
+
+%bestTranslations = ();
+while($line = <VOTES>) {
+    chomp $line;
+    @fields = split(/\t/, $line);
+    $doc_seg_ids = $fields[0];
+    ($docID, $segID) = split(/_/, $doc_seg_ids);
+    $bestTranslation = $fields[2];
+    $bestTranslations{$docID}{$segID} = $bestTranslation;
+}
+close(VOTES);
+
+%sourceSentences = ();
+while($line = <SEGID_SOURCE>) {
+    chomp $line;
+    @fields = split(/\t/, $line);
+    $doc_seg_ids = $fields[0];
+    ($docID, $segID) = split(/_/, $doc_seg_ids);
+    $source = $fields[1];
+    $sourceSentences{$docID}{$segID} = $source;
+}
+close(SEGID_SOURCE);
+
+foreach $docID (sort { $a <=> $b } keys %bestTranslations) {
+    foreach $segID (sort { $a <=> $b } keys %{$bestTranslations{$docID}}) {
+	if($sourceSentences{$docID}{$segID}) {
+	    print $docID . "\t" . $segID . "\t" . $sourceSentences{$docID}{$segID} . "\t" . $bestTranslations{$docID}{$segID} . "\n";
+	}
+    }
+}