From 0afd9d510cde40c340cf2c389b1aa22b5a9379c5 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')

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