From 3a162d28033d1b9d5241e31f32978dba4eba6296 Mon Sep 17 00:00:00 2001
From: Chris Dyer <cdyer@allegro.clab.cs.cmu.edu>
Date: Fri, 8 Mar 2013 22:44:49 -0500
Subject: few preproc fixes

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

(limited to 'corpus')

diff --git a/corpus/paste-files.pl b/corpus/paste-files.pl
index 4cb424ad..ef2cd937 100755
--- a/corpus/paste-files.pl
+++ b/corpus/paste-files.pl
@@ -34,6 +34,7 @@ while(1) {
       $done = 1;
       last;
     }
+    $r =~ s/\r//g;
     chomp $r;
     if ($r =~ /\|\|\|/) {
       $r = '';
diff --git a/corpus/support/quote-norm.pl b/corpus/support/quote-norm.pl
index d2980092..b104e73c 100755
--- a/corpus/support/quote-norm.pl
+++ b/corpus/support/quote-norm.pl
@@ -11,6 +11,8 @@ while(<STDIN>) {
   s/&\s*squot\s*;/'/gi;
   s/&\s*quot\s*;/"/gi;
   s/&\s*amp\s*;/&/gi;
+  s/&\s*nbsp\s*;/&/gi;
+  s/&\s*#\s*160\s*;/ /gi;
   s/ (\d\d): (\d\d)/ $1:$2/g;
   s/[\x{20a0}]\x{20ac}]/ EUR /g;
   s/[\x{00A3}]/ GBP /g;
diff --git a/corpus/support/token_list b/corpus/support/token_list
index d470cb22..366cd7ff 100644
--- a/corpus/support/token_list
+++ b/corpus/support/token_list
@@ -37,6 +37,8 @@ tel.
 10.
 
 ##################### abbreviation: words that contain period.
+EE.UU.
+ee.uu.
 U.A.E
 Ala.
 Ph.D.
-- 
cgit v1.2.3