| 1 |
#!/usr/bin/perl -w |
package StemHr; |
| 2 |
|
|
| 3 |
# |
# |
| 4 |
# Croatian stemmer |
# Croatian stemmer |
| 5 |
# |
# |
| 19 |
# - uzvici |
# - uzvici |
| 20 |
# |
# |
| 21 |
|
|
| 22 |
sub rm_a { |
use strict; |
| 23 |
s/a//g; |
use locale; |
| 24 |
} |
|
| 25 |
|
use Memoize; |
| 26 |
|
#memoize('stem'); |
| 27 |
|
|
| 28 |
sub kgh { |
sub kgh { |
| 29 |
my ($pre,$replace,$post) = @_; |
my ($pre,$replace,$post) = @_; |
| 30 |
$replace =~ s/[cè]/k/g; |
$replace =~ s/[cè]/k/g; |
| 31 |
$replace =~ s/[z¾]/g/g; |
$replace =~ s/[z¾]/g/g; |
| 32 |
$replace =~ s/[s¹]/h/g; |
$replace =~ s/[s¹]/h/g; |
| 33 |
return $pre . $replace . $post . "\tkgh"; |
return $pre . $replace . $post; |
| 34 |
} |
} |
| 35 |
|
|
| 36 |
# samoglasnici |
# samoglasnici |
| 38 |
# suglasnici |
# suglasnici |
| 39 |
my $su = '[^aeiou]'; |
my $su = '[^aeiou]'; |
| 40 |
|
|
| 41 |
while(<>) { |
my $palatal = '(lj|nj|j|æ|¾|¹|¾|¹t|¾d)'; |
|
chomp; |
|
| 42 |
|
|
| 43 |
my $orig = $_; |
# glagolni razredi |
| 44 |
|
my $g_1r = '[td]'; |
| 45 |
|
my $g_2r = '[sz]'; |
| 46 |
|
my $g_3r = '[pb]'; |
| 47 |
|
my $g_4r = '[kgh]'; |
| 48 |
|
my $g_5r = '[nm]'; |
| 49 |
|
my $g_6r = '(:?nu|n)'; |
| 50 |
|
|
| 51 |
# imenice |
sub stem { |
| 52 |
# vrsta a |
|
| 53 |
|
my $w = shift || return; |
| 54 |
|
|
| 55 |
unless ( |
unless ( |
| 56 |
|
# infinitiv |
| 57 |
|
$w =~ s/(\w)(ti|æi)$/$1.$2 100/g || |
| 58 |
|
# 1. razred |
| 59 |
|
$w =~ s/([^sk])[td](em|e¹|e|emo|ete|oh|osmo|oste|o¹e|ijah|ija¹e|ijasno|ijaste|ijahu|imo|ite|en|ena|eni)$/$1s.ti 101/g || |
| 60 |
|
# 2. razred |
| 61 |
|
$w =~ s/(${sa})[sz](em|e¹|e|e¹emo|emo|ete|u|oh|e|osmo|oste|o¹e|ijah|ija¹e|ijasmo|ijaste|ijahu|imo|ite|uæi|av¹i|ao|la|lo|li|le|la|en|ena|eni)$/$1s.ti 102/ || |
| 62 |
|
# 3. razred |
| 63 |
|
$w =~ s/(p|b|sp|su)(em|e¹|e|emo|ete|u|oh|osmo|oste|o¹e|ah|a¹e|asmo|aste|ahu|ijah|ija¹e|ijahu|i|imo|ite||uæi|av¹i|ao|la|lo|en|ena|eni)$/$1s.ti 103/ || |
| 64 |
|
# 4. razred |
| 65 |
|
$w =~ s/[è¾¹](em|e¹|e|emo|ete|u|ah|ahu|en|ena)$/.æi 104/g || |
| 66 |
|
$w =~ s/[k](oh|osmo|oste|o¹e|uæi|av¹i|ao|la|lo)$/.æi 105/g || |
| 67 |
|
$w =~ s/[c](ijah|ija¹e|ijasmo|ijaste|ijahu|i|imo|ite)$/.æi 106/g || |
| 68 |
|
$w =~ s/[g](nuti|oh|nuh|nu|av¹i|nuv¹i|ao|nuo|nem|ne¹|ni|imo|nut|nimo)$/.æi 107/g || |
| 69 |
|
|
| 70 |
|
# imenice |
| 71 |
|
|
| 72 |
|
# vrsta a |
| 73 |
|
$w =~ s/(${su}st)a$/$1 13/g || |
| 74 |
|
$w =~ s/(${su})c[ae]$/$1ce 17/g || |
| 75 |
|
# kgh |
| 76 |
|
$w =~ s/(\w${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,' 1')/gex || |
| 77 |
|
$w =~ s/(${sa}[kgh])(a|u|om)$/$1 2/g || |
| 78 |
|
$w =~ s/(${su})([è¾¹czs])(i|e|ima)$/kgh($1,"a$2",' 3')/gex || |
| 79 |
|
# imenice na palatal |
| 80 |
|
$w =~ s/${palatal}${palatal}(a|u|em|i|ima|e)$/$1a$2 9/g || |
| 81 |
|
$w =~ s/${palatal}(a|u|em)$/$1 8/g || |
| 82 |
# nepostojano a |
# nepostojano a |
| 83 |
s/(${su}a{$su})a/$1 1/g || |
$w =~ s/(${su}a{$su})a/$1 4/g || |
| 84 |
s/(${su})(${su})(a|u|i|e|om|ima)$/$1a$2\t3/g |
$w =~ s/(${su})(${su})(a|u|i|e|om|ima)$/$1a$2 5/g |
| 85 |
) { |
) { |
| 86 |
# kgh |
|
| 87 |
s/(${su}${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,'')/gex; |
# vrsta a |
| 88 |
# s/${su}${sa}[c](i|ima) 4//g; |
$w =~ s/me$/me 11/g || |
| 89 |
s/(${su}${sa}[kgh])(a|u|om)\t5/$1/g; |
$w =~ s/(eta|ena)$/e 12/g || |
| 90 |
|
|
| 91 |
# mno¾ina |
$w =~ s/(\w${sa}${su})(o|e|a|u|om|em|i|ima|ina|eta)$/$1 7/g || |
| 92 |
s/(${su})(ov|ev)*(i|a|ima|e|in|ina)$/$1\t6/g; |
|
| 93 |
# jednina |
$w =~ s/(${su})sa$/$1as 14/g || |
| 94 |
s/(o|e|a|u|om|em)$/\t7/g; |
$w =~ s/(${su})ena$/$1e 16/g || |
| 95 |
|
$w =~ s/(${su})eta$/$1e 17/g || |
| 96 |
|
$w =~ s/(${su})([oe])$/$1$2 18/g || |
| 97 |
|
|
| 98 |
|
0; |
| 99 |
|
|
| 100 |
} |
} |
| 101 |
|
|
| 102 |
|
# makni broj pravila |
| 103 |
|
$w =~ s/\s\d+$//; |
| 104 |
|
# makni toèku koja oznaèava korjen rijeèi |
| 105 |
|
$w =~ s/\.//g; |
| 106 |
|
|
| 107 |
printf("%-15s %s\n",$orig,$_); |
return $w; |
| 108 |
} |
} |
| 109 |
|
|
Parent Directory
| 
Revision Log
| 
Patch