From 284383880f043edb2d67afbe2f64237c466245c1 Mon Sep 17 00:00:00 2001
From: Chris Dyer <cdyer@allegro.clab.cs.cmu.edu>
Date: Mon, 10 Mar 2014 18:40:13 -0400
Subject: few tokenization bugs

---
 corpus/support/quote-norm.pl | 1 +
 corpus/support/token_list    | 6 ++++--
 2 files changed, 5 insertions(+), 2 deletions(-)

(limited to 'corpus/support')

diff --git a/corpus/support/quote-norm.pl b/corpus/support/quote-norm.pl
index 33604027..0366fad5 100755
--- a/corpus/support/quote-norm.pl
+++ b/corpus/support/quote-norm.pl
@@ -39,6 +39,7 @@ while(<STDIN>) {
   s/&\#([0-9]+);/pack("U", $1)/ge;
 
   # Regularlize spaces:
+  s/\x{ad}//g;        # soft hyphen
   s/\x{a0}/ /g;       # non-breaking space
   s/\x{2009}/ /g;     # thin space
   s/\x{2028}/ /g;     # "line separator"
diff --git a/corpus/support/token_list b/corpus/support/token_list
index 228663f6..d38638cf 100644
--- a/corpus/support/token_list
+++ b/corpus/support/token_list
@@ -156,8 +156,9 @@ Mass.
 Md.
 Mfg.
 Mgr.
-Mexican-U.S.
-Mich.
+Mio.
+Mrd.
+Bio.
 Minn.
 Mo.
 Mon.
@@ -187,6 +188,7 @@ Rd.
 Rev.
 R.J.
 C.L
+Rs.
 Rte.
 Sat.
 W.T
-- 
cgit v1.2.3


From b2b25c9e71a7ae1b20b7350a236419aa37179e7c Mon Sep 17 00:00:00 2001
From: mjdenkowski <michael.j.denkowski@gmail.com>
Date: Wed, 12 Mar 2014 22:06:35 -0400
Subject: XML file tokenization for all your WMT needs.

---
 corpus/support/fix-eos.pl |  2 ++
 corpus/xml-tok.py         | 47 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 49 insertions(+)
 create mode 100755 corpus/xml-tok.py

(limited to 'corpus/support')

diff --git a/corpus/support/fix-eos.pl b/corpus/support/fix-eos.pl
index 584f8b46..fe03727b 100755
--- a/corpus/support/fix-eos.pl
+++ b/corpus/support/fix-eos.pl
@@ -1,4 +1,6 @@
 #!/usr/bin/perl -w
+$|++;
+
 use strict;
 use utf8;
 
diff --git a/corpus/xml-tok.py b/corpus/xml-tok.py
new file mode 100755
index 00000000..4357ced6
--- /dev/null
+++ b/corpus/xml-tok.py
@@ -0,0 +1,47 @@
+#!/usr/bin/env python
+
+import os
+import re
+import subprocess
+import sys
+
+# Tokenize XML files with tokenize-anything.sh
+# in:  <seg id="963"> The earnings on its 10-year bonds are 28.45%. </seg>
+# out: <seg id="963"> The earnings on its 10 - year bonds are 28.45 % . </seg>
+
+def escape(s):
+    return s.replace('&', '&amp;').replace('>', '&gt;').replace('<', '&lt;').replace('"', '&quot;').replace('\'', '&apos;')
+
+def unescape(s):
+    return s.replace('&gt;', '>').replace('&lt;', '<').replace('&quot;', '"').replace('&apos;', '\'').replace('&amp;', '&')
+
+def main():
+    tok = subprocess.Popen([os.path.join(os.path.dirname(__file__), 'tokenize-anything.sh'), '-u'], stdin=subprocess.PIPE, stdout=subprocess.PIPE)
+    while True:
+        line = sys.stdin.readline()
+        if not line:
+            break
+        line = line.strip()
+        pieces = []
+        eol = len(line)
+        pos = 0
+        while pos < eol:
+            next = line.find('<', pos)
+            if next == -1:
+                next = eol
+            tok.stdin.write('{}\n'.format(unescape(line[pos:next])))
+            pieces.append(escape(tok.stdout.readline().strip()))
+            if next == eol:
+                break
+            pos = line.find('>', next + 1)
+            if pos == -1:
+                pos = eol
+            else:
+                pos += 1
+            pieces.append(line[next:pos])
+        sys.stdout.write('{}\n'.format(' '.join(pieces).strip()))
+    tok.stdin.close()
+    tok.wait()
+
+if __name__ == '__main__':
+    main()
-- 
cgit v1.2.3