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 | 5b8253e0e1f1393a509fb9975ba8c1347af758ed (patch) | |
| tree | 1790470b1d07a0b4973ebce19192e896566ea60b /corpus/tokenize-anything.sh | |
| parent | 2389a5a8a43dda87c355579838559515b0428421 (diff) | |
| parent | b203f8c5dc8cff1b9c9c2073832b248fcad0765a (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;}' + |