diff options
| author | Chris Dyer <cdyer@allegro.clab.cs.cmu.edu> | 2014-12-29 19:20:49 -0500 |
|---|---|---|
| committer | Chris Dyer <cdyer@allegro.clab.cs.cmu.edu> | 2014-12-29 19:20:49 -0500 |
| commit | 992556cc0931b255b9e299c0f489c3b449b22ab4 (patch) | |
| tree | 8251b4d8fa53464369c6da443d919922d92dd6b6 | |
| parent | a66b19342f105546adfb9cd9189f862e8a9c0b15 (diff) | |
deal with eur symbol
| -rwxr-xr-x | corpus/support/tokenizer.pl | 8 |
1 files changed, 4 insertions, 4 deletions
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){ |