From 9f57df1eb1346196b6ea755eab9fb81a8a57b915 Mon Sep 17 00:00:00 2001 From: Chris Dyer Date: Wed, 15 Jan 2014 23:33:43 -0500 Subject: deal with hindi --- corpus/support/tokenizer.pl | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) diff --git a/corpus/support/tokenizer.pl b/corpus/support/tokenizer.pl index e0df16a7..079f45b6 100755 --- a/corpus/support/tokenizer.pl +++ b/corpus/support/tokenizer.pl @@ -65,7 +65,7 @@ my $Split_AposD = 1; ## 'd ### some patterns -my $common_right_punc = '\.|\,|\;|:|\!|\?|\"|\)|\]|\}|\>|\-'; +my $common_right_punc = '\x{0964}|\.|\,|\;|\!|:|\?|\"|\)|\]|\}|\>|\-'; #### step 1: read files @@ -112,7 +112,6 @@ my $new_token_total = 0; while(){ chomp(); - if(/^(\[b\s+|\]b|\]f|\[f\s+)/ || (/^\[[bf]$/) || (/^\s*$/) || /^){ my $orig_num = 0; my $deep_proc_num = 0; - + s/(\x{0964}+)/ $1/g; # Devangari end of sentence my $new_line = proc_line($_, \$orig_num, \$deep_proc_num); $orig_token_total += $orig_num; @@ -148,7 +147,8 @@ while(){ $new_line =~ s/(set|src|tgt|trg)/ $1/g; } - print STDOUT " $new_line\n"; + chomp $new_line; + print STDOUT "$new_line\n"; } ######################################################################## @@ -228,6 +228,7 @@ sub proc_token { ## step 1: check the most common case if($token =~ /^[a-z0-9\p{Cyrillic}\p{Greek}\p{Hebrew}\p{Han}\p{Arabic}\p{Devanagari}]+$/i){ + #if($token =~ /^[a-z0-9\p{Cyrillic}\p{Greek}\p{Hebrew}\p{Han}\p{Arabic}]+$/i){ ### most common cases return $token; } @@ -363,7 +364,7 @@ sub deep_proc_token { ##### step 0: if it mades up of all puncts, remove one punct at a time. if($line !~ /[\p{Cyrillic}\p{Greek}\p{Hebrew}\p{Han}\p{Arabic}\p{Devanagari}a-zA-Z\d]/){ - if($line =~ /^(\!+|\@+|\++|\=+|\*+|\<+|\>+|\|+|\?+|\.+|\-+|\_+|\&+)$/){ + if($line =~ /^(\!+|\@+|\++|\=+|\*+|\<+|\>+|\|+|\?+|\x{0964}+|\.+|\-+|\_+|\&+)$/){ ## ++ @@@@ !!! .... return $line; } @@ -454,7 +455,7 @@ sub deep_proc_token { ### deal with ': e.g., 's, 't, 'm, 'll, 're, 've, n't ## 'there => ' there '98 => the same - $suc += ($line =~ s/^(\'+)([a-z]+)/ $1 $2/gi); + $suc += ($line =~ s/^(\'+)([a-z\p{Cyrillic}\p{Greek}\p{Hebrew}\p{Han}\p{Arabic}\p{Devanagari}]+)/ $1 $2/gi); ## note that \' and \. could interact: e.g., U.S.'s; 're. if($Split_NAposT && ($line =~ /^(.*[a-z]+)(n\'t)([\.]*)$/i)){ -- cgit v1.2.3 From ef3c7beb9e58f22df5a778deed185d751dda0580 Mon Sep 17 00:00:00 2001 From: Chris Dyer Date: Thu, 16 Jan 2014 00:25:14 -0500 Subject: moar hindi --- corpus/support/quote-norm.pl | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/corpus/support/quote-norm.pl b/corpus/support/quote-norm.pl index 57f4ad77..7fe33db4 100755 --- a/corpus/support/quote-norm.pl +++ b/corpus/support/quote-norm.pl @@ -61,6 +61,11 @@ while() { s/«/"/g; s/»/"/g; tr/!-~/!-~/; + tr/०-९/0-9/; # devangari + tr/౦-౯/0-9/; # telugu + tr/೦-೯/0-9/; # kannada + tr/೦-௯/0-9/; # tamil + tr/൦-൯/0-9/; # malayalam s/、/,/g; # s/。/./g; s/…/.../g; -- cgit v1.2.3