Pipeline code for running with mixing tokens and tags in the clustering.

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@377 ec762483-ff6d-05da-a07a-a48fb63a330f

3 files changed, 99 insertions, 36 deletions

diff --git a/gi/pipeline/scripts/patch-corpus.pl b/gi/pipeline/scripts/patch-corpus.pl
index 200022bc..c0eec43e 100755
--- a/gi/pipeline/scripts/patch-corpus.pl
+++ b/gi/pipeline/scripts/patch-corpus.pl
@@ -3,12 +3,17 @@ use strict;
 
 my $PATCH = shift @ARGV;
 my $TGT = 1;
-if ($PATCH eq "-s") {
-    undef $TGT;
+my $APPEND;
+while ($PATCH eq "-s" || $PATCH eq "-a") {
+    if ($PATCH eq "-s") {
+        undef $TGT;
+    } else {
+        $APPEND = 1;
+    }
     $PATCH = shift @ARGV;
 }
 
-die "Usage: $0 [-s] tagged.en[_fr] < lexical.en_fr_al[_...]\n" unless $PATCH;
+die "Usage: $0 [-s] [-a] tagged.en[_fr] < lexical.en_fr_al[_...]\n" unless $PATCH;
 
 open P, "<$PATCH" or die "Can't read tagged corpus $PATCH: $!";
 my $first=<P>; close P;
@@ -33,11 +38,25 @@ while(my $pline = <P>) {
   if ($TGT) {
       my @lwords = split /\s+/, $fields[1];
       die "Length mismatch in line $line!\n" unless (scalar @pwords == scalar @lwords);
-      $fields[1] = $pline;
-  } else {
+      if ($APPEND) {
+          foreach my $i (0..(scalar @pwords-1)) {
+              $lwords[$i] = $lwords[$i] . '_' . $pwords[$i];
+          }
+          $fields[1] = join ' ', @lwords;
+      } else {
+          $fields[1] = $pline;
+      }
+  } else { # source side
       my @lwords = split /\s+/, $fields[0];
       die "Length mismatch in line $line!\n" unless (scalar @pwords == scalar @lwords);
-      $fields[0] = $pline;
+      if ($APPEND) {
+          foreach my $i (0..(scalar @pwords-1)) {
+              $lwords[$i] = $lwords[$i] . '_' . $pwords[$i];
+          }
+          $fields[0] = join ' ', @lwords;
+      } else {
+          $fields[0] = $pline;
+      }
   }
   print join ' ||| ', @fields;
   print "\n";
diff --git a/gi/pipeline/scripts/remove-tags-from-contexts.pl b/gi/pipeline/scripts/remove-tags-from-contexts.pl
new file mode 100755
index 00000000..20698816
--- /dev/null
+++ b/gi/pipeline/scripts/remove-tags-from-contexts.pl
@@ -0,0 +1,53 @@
+#!/usr/bin/perl -w
+use strict;
+
+use Getopt::Long "GetOptions";
+
+my $PHRASE = 'tok';
+my $CONTEXT = 'tag';
+
+die "Usage: $0 [--phrase=tok|tag] [--context=tok|tag] < corpus" 
+    unless &GetOptions('phrase=s' => \$PHRASE, 'context=s' => \$CONTEXT);
+
+my $lno = 0;
+while(my $line = <>) {
+    $lno++;
+    chomp $line;
+    my @top = split /\t/, $line;
+    die unless (scalar @top == 2); 
+
+    my @pwords = split /\s+/, $top[0];
+    foreach my $token (@pwords) {
+        #print $token . "\n";
+        my @parts = split /_(?!.*_)/, $token;
+        die unless (scalar @parts == 2); 
+        if ($PHRASE eq "tok") {
+            $token = $parts[0]
+        } elsif ($PHRASE eq "tag") {
+            $token = $parts[1]
+        }
+    }
+
+    my @fields = split / \|\|\| /, $top[1];
+    foreach my $i (0..((scalar @fields) / 2 - 1)) {
+        #print $i . ": " . $fields[2*$i] . " of " . (scalar @fields) . "\n";
+        my @cwords = split /\s+/, $fields[2*$i];
+        foreach my $token (@cwords) {
+            #print $i . ": " . $token . "\n";
+            my @parts = split /_(?!.*_)/, $token;
+            if (scalar @parts == 2) {
+                if ($CONTEXT eq "tok") {
+                    $token = $parts[0]
+                } elsif ($CONTEXT eq "tag") {
+                    $token = $parts[1]
+                }
+            }
+        }
+        $fields[2*$i] = join ' ', @cwords;
+    }
+
+    print join ' ', @pwords;
+    print "\t";
+    print join ' ||| ', @fields;
+    print "\n";
+}
diff --git a/gi/pipeline/scripts/remove-tags-from-corpus.pl b/gi/pipeline/scripts/remove-tags-from-corpus.pl
index 5460db95..be3e97c0 100755
--- a/gi/pipeline/scripts/remove-tags-from-corpus.pl
+++ b/gi/pipeline/scripts/remove-tags-from-corpus.pl
@@ -3,51 +3,42 @@ use strict;
 
 use Getopt::Long "GetOptions";
 
-my $PHRASE = 'tok';
-my $CONTEXT = 'tag';
-
-die "Usage: $0 [--phrase=tok|tag] [--context=tok|tag] < corpus" 
-    unless &GetOptions('phrase=s' => \$PHRASE, 'context=s' => \$CONTEXT);
+my $LANGUAGE = shift @ARGV;
+$LANGUAGE = 'target' unless ($LANGUAGE);
 
 my $lno = 0;
 while(my $line = <>) {
     $lno++;
     chomp $line;
-    my @top = split /\t/, $line;
-    die unless (scalar @top == 2); 
 
-    my @pwords = split /\s+/, $top[0];
-    foreach my $token (@pwords) {
-        #print $token . "\n";
-        my @parts = split /_(?!_)/, $token;
-        die unless (scalar @parts == 2); 
-        if ($PHRASE eq "tok") {
-            $token = $parts[0]
-        } elsif ($PHRASE eq "tag") {
-            $token = $parts[1]
+    my @fields = split / \|\|\| /, $line;
+
+    if ($LANGUAGE eq "source" or $LANGUAGE eq "both") {
+        my @cwords = split /\s+/, $fields[0];
+        foreach my $token (@cwords) {
+            my @parts = split /_(?!.*_)/, $token;
+            if (scalar @parts == 2) {
+                $token = $parts[0]
+            } else {
+                print STDERR "WARNING: invalid tagged token $token\n";
+            }
         }
+        $fields[0] = join ' ', @cwords;
     }
 
-    my @fields = split / \|\|\| /, $top[1];
-    foreach my $i (0..((scalar @fields) / 2 - 1)) {
-        #print $i . ": " . $fields[2*$i] . " of " . (scalar @fields) . "\n";
-        my @cwords = split /\s+/, $fields[2*$i];
+    if ($LANGUAGE eq "target" or $LANGUAGE eq "both") {
+        my @cwords = split /\s+/, $fields[1];
         foreach my $token (@cwords) {
-            #print $i . ": " . $token . "\n";
-            my @parts = split /_/, $token;
+            my @parts = split /_(?!.*_)/, $token;
             if (scalar @parts == 2) {
-                if ($CONTEXT eq "tok") {
-                    $token = $parts[0]
-                } elsif ($CONTEXT eq "tag") {
-                    $token = $parts[1]
-                }
+                $token = $parts[1]
+            } else {
+                print STDERR "WARNING: invalid tagged token $token\n";
             }
         }
-        $fields[2*$i] = join ' ', @cwords;
+        $fields[0] = join ' ', @cwords;
     }
 
-    print join ' ', @pwords;
-    print "\t";
     print join ' ||| ', @fields;
     print "\n";
 }