diff options
Diffstat (limited to 'corpus/support')
| -rwxr-xr-x | corpus/support/quote-norm.pl | 1 | ||||
| -rwxr-xr-x | corpus/support/tokenizer.pl | 4 |
2 files changed, 5 insertions, 0 deletions
diff --git a/corpus/support/quote-norm.pl b/corpus/support/quote-norm.pl index 0366fad5..3eee0666 100755 --- a/corpus/support/quote-norm.pl +++ b/corpus/support/quote-norm.pl @@ -40,6 +40,7 @@ while(<STDIN>) { # Regularlize spaces: s/\x{ad}//g; # soft hyphen + s/\x{200C}//g; # zero-width non-joiner s/\x{a0}/ /g; # non-breaking space s/\x{2009}/ /g; # thin space s/\x{2028}/ /g; # "line separator" diff --git a/corpus/support/tokenizer.pl b/corpus/support/tokenizer.pl index 7771201f..f57bc87a 100755 --- a/corpus/support/tokenizer.pl +++ b/corpus/support/tokenizer.pl @@ -240,6 +240,10 @@ sub proc_token { return $token; } + if($token =~ /^\d+(.\d+)+(亿|百万|万|千)?$/){ + return $token; + } + ## 1,234,345.34 if($token =~ /^\d+(\.\d{3})*,\d+$/){ ## number |