diff options
Diffstat (limited to 'corpus/support')
| -rw-r--r-- | corpus/support/token_list | 49 | ||||
| -rw-r--r-- | corpus/support/token_patterns | 2 | ||||
| -rwxr-xr-x | corpus/support/tokenizer.pl | 8 |
3 files changed, 55 insertions, 4 deletions
diff --git a/corpus/support/token_list b/corpus/support/token_list index d38638cf..00daa82b 100644 --- a/corpus/support/token_list +++ b/corpus/support/token_list @@ -1,6 +1,55 @@ ##################### hyphenated words added by Fei since 3/7/05 ##X-ray +# Finnish +eaa. +ap. +arv. +ay. +eKr. +em. +engl. +esim. +fil. +lis. +fil. +maist. +fil.toht. +harv. +ilt. +jatk. +jKr. +jms. +jne. +joht. +klo +ko. +ks. +leht. +lv. +lyh. +mm. +mon. +nim. +nro. +ns. +nti. +os. +oy. +pj. +pnä. +puh. +pvm. +rva. +tms. +ts. +vars. +vrt. +ym. +yms. +yo. +>>>>>>> 8646b68e5b124f612fd65b51ea40624f65a2f3d6 + # hindi abbreviation patterns जन. फर. diff --git a/corpus/support/token_patterns b/corpus/support/token_patterns index de64fb2a..12558cdd 100644 --- a/corpus/support/token_patterns +++ b/corpus/support/token_patterns @@ -1,4 +1,6 @@ /^(al|el|ul|e)\-[a-z]+$/ +/\.(fi|fr|es|co\.uk|de)$/ +/:[a-zä]+$/ /^((а|А)(ль|ш)|уль)-\p{Cyrillic}+$/ /^\p{Cyrillic}\.\p{Cyrillic}\.$/ /^(\d|\d\d|\d\d\d)\.$/ 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){ |