diff options
| author | Avneesh Saluja <asaluja@gmail.com> | 2013-03-28 18:28:16 -0700 |
|---|---|---|
| committer | Avneesh Saluja <asaluja@gmail.com> | 2013-03-28 18:28:16 -0700 |
| commit | 3d8d656fa7911524e0e6885647173474524e0784 (patch) | |
| tree | 81b1ee2fcb67980376d03f0aa48e42e53abff222 /corpus/tokenize-anything.sh | |
| parent | be7f57fdd484e063775d7abf083b9fa4c403b610 (diff) | |
| parent | 96fedabebafe7a38a6d5928be8fff767e411d705 (diff) | |
fixed conflicts
Diffstat (limited to 'corpus/tokenize-anything.sh')
| -rwxr-xr-x | corpus/tokenize-anything.sh | 13 |
1 files changed, 13 insertions, 0 deletions
diff --git a/corpus/tokenize-anything.sh b/corpus/tokenize-anything.sh new file mode 100755 index 00000000..028992cf --- /dev/null +++ b/corpus/tokenize-anything.sh @@ -0,0 +1,13 @@ +#!/bin/sh + +ROOTDIR=`dirname $0` +SUPPORT=$ROOTDIR/support + +$SUPPORT/utf8-normalize.sh | + $SUPPORT/quote-norm.pl | + $SUPPORT/tokenizer.pl | + sed -e 's/ al - / al-/g' | + $SUPPORT/fix-contract.pl | + sed -e 's/^ //' | sed -e 's/ $//' | + perl -e 'while(<>){s/(\d+)(\.+)$/$1 ./; s/(\d+)(\.+) \|\|\|/$1 . |||/; print;}' + |