Merge branch 'master' of github.com:redpony/cdec

1 files changed, 25 insertions, 0 deletions

diff --git a/corpus/utf8-normalize.sh b/corpus/utf8-normalize.sh
new file mode 100755
index 00000000..dcf8bc59
--- /dev/null
+++ b/corpus/utf8-normalize.sh
@@ -0,0 +1,25 @@
+#!/bin/bash
+
+# This script uses ICU uconv (http://site.icu-project.org/), if it's available
+# to normalize UTF8 text into a standard form. For information about this
+# process, refer to http://en.wikipedia.org/wiki/Unicode_equivalence#Normalization
+# Escape characters between 0x00-0x1F are removed
+
+if which uconv > /dev/null
+then
+  CMD="uconv -f utf8 -t utf8 -x Any-NFKC --callback skip"
+else
+  echo "Cannot find ICU uconv (http://site.icu-project.org/) ... falling back to iconv. Normalization NOT taking place." 1>&2
+  CMD="iconv -f utf8 -t utf8 -c"
+fi
+
+$CMD | /usr/bin/perl -w -e '
+ while (<>) {
+     chomp;
+      s/[\x00-\x1F]+/ /g;
+      s/  +/ /g;
+      s/^ //;
+      s/ $//;
+      print "$_\n";
+    }'
+