From 379284b264f3246b2849369bfec7bb16cc701af3 Mon Sep 17 00:00:00 2001 From: Chris Dyer Date: Mon, 29 Dec 2014 19:20:49 -0500 Subject: deal with eur symbol --- corpus/support/tokenizer.pl | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'corpus/support') diff --git a/corpus/support/tokenizer.pl b/corpus/support/tokenizer.pl index aa285be4..718d78cc 100755 --- a/corpus/support/tokenizer.pl +++ b/corpus/support/tokenizer.pl @@ -415,7 +415,7 @@ sub deep_proc_token { } ## remove the ending periods that follow number etc. - if($line =~ /^(.*(\d|\~|\^|\&|\:|\,|\#|\*|\%|\-|\_|\/|\\|\$|\'))(\.+)$/){ + if($line =~ /^(.*(\d|\~|\^|\&|\:|\,|\#|\*|\%|€|\-|\_|\/|\\|\$|\'))(\.+)$/){ ## 12~13. => 12~13 . my $t1 = $1; my $t3 = $3; @@ -600,12 +600,12 @@ sub deep_proc_token { ## deal with "%" - if(($line =~ /\%/) && ($Split_On_PercentSign > 0)){ + if(($line =~ /\%|€/) && ($Split_On_PercentSign > 0)){ my $suc = 0; if($Split_On_PercentSign >= 2){ - $suc += ($line =~ s/(\D)(\%+)/$1 $2/g); + $suc += ($line =~ s/(\D)(\%+|€+)/$1 $2/g); }else{ - $suc += ($line =~ s/(\%+)/ $1 /g); + $suc += ($line =~ s/(\%+|€+)/ $1 /g); } if($suc){ -- cgit v1.2.3