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 |
sub kgh { |
sub kgh { |
26 |
my ($pre,$replace,$post) = @_; |
my ($pre,$replace,$post) = @_; |
27 |
$replace =~ s/[cè]/k/g; |
$replace =~ s/[cè]/k/g; |
28 |
$replace =~ s/[z¾]/g/g; |
$replace =~ s/[z¾]/g/g; |
29 |
$replace =~ s/[s¹]/h/g; |
$replace =~ s/[s¹]/h/g; |
30 |
return $pre . $replace . $post . "\tkgh"; |
return $pre . $replace . $post; |
31 |
} |
} |
32 |
|
|
33 |
# samoglasnici |
# samoglasnici |
35 |
# suglasnici |
# suglasnici |
36 |
my $su = '[^aeiou]'; |
my $su = '[^aeiou]'; |
37 |
|
|
38 |
while(<>) { |
my $palatal = '(lj|nj|j|æ|¾|¹|¾|¹t|¾d)'; |
|
chomp; |
|
39 |
|
|
40 |
my $orig = $_; |
# glagolni razredi |
41 |
|
my $g_1r = '[td]'; |
42 |
|
my $g_2r = '[sz]'; |
43 |
|
my $g_3r = '[pb]'; |
44 |
|
my $g_4r = '[kgh]'; |
45 |
|
my $g_5r = '[nm]'; |
46 |
|
my $g_6r = '(:?nu|n)'; |
47 |
|
|
48 |
# imenice |
sub stem { |
|
# vrsta a |
|
49 |
|
|
50 |
unless ( |
unless ( |
51 |
|
# infinitiv |
52 |
|
s/(\w)(ti|æi)$/$1.$2 100/g || |
53 |
|
# 1. razred |
54 |
|
s/([^sk])[td](em|e¹|e|emo|ete|oh|osmo|oste|o¹e|ijah|ija¹e|ijasno|ijaste|ijahu|imo|ite|en|ena|eni)$/$1s.ti 101/g || |
55 |
|
# 2. razred |
56 |
|
s/(${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)$/$1s.ti 102/ || |
57 |
|
# 3. razred |
58 |
|
s/(p|b|sp|su)(em|e¹|e|emo|ete|u|oh|osmo|oste|o¹e|ah|a¹e|asmo|aste|ahu|ijah|ija¹e|ijahu|i|imo|ite||uæi|av¹i|ao|la|lo|en|ena|eni)$/$1s.ti 103/ || |
59 |
|
# 4. razred |
60 |
|
s/[è¾¹](em|e¹|e|emo|ete|u|ah|ahu|en|ena)$/.æi 104/g || |
61 |
|
s/[k](oh|osmo|oste|o¹e|uæi|av¹i|ao|la|lo)$/.æi 105/g || |
62 |
|
s/[c](ijah|ija¹e|ijasmo|ijaste|ijahu|i|imo|ite)$/.æi 106/g || |
63 |
|
s/[g](nuti|oh|nuh|nu|av¹i|nuv¹i|ao|nuo|nem|ne¹|ni|imo|nut|nimo)$/.æi 107/g || |
64 |
|
|
65 |
|
# imenice |
66 |
|
|
67 |
|
# vrsta a |
68 |
|
s/(${su}st)a$/$1 13/g || |
69 |
|
s/(${su})c[ae]$/$1ce 17/g || |
70 |
|
# kgh |
71 |
|
s/(\w${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,' 1')/gex || |
72 |
|
s/(${sa}[kgh])(a|u|om)$/$1 2/g || |
73 |
|
s/(${su})([è¾¹czs])(i|e|ima)$/kgh($1,"a$2",' 3')/gex || |
74 |
|
# imenice na palatal |
75 |
|
s/${palatal}${palatal}(a|u|em|i|ima|e)$/$1a$2 9/g || |
76 |
|
s/${palatal}(a|u|em)$/$1 8/g || |
77 |
# nepostojano a |
# nepostojano a |
78 |
s/(${su}a{$su})a/$1 1/g || |
s/(${su}a{$su})a/$1 4/g || |
79 |
s/(${su})(${su})(a|u|i|e|om|ima)$/$1a$2\t3/g |
s/(${su})(${su})(a|u|i|e|om|ima)$/$1a$2 5/g |
80 |
) { |
) { |
|
# kgh |
|
|
s/(${su}${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,'')/gex; |
|
|
# s/${su}${sa}[c](i|ima) 4//g; |
|
|
s/(${su}${sa}[kgh])(a|u|om)\t5/$1/g; |
|
|
|
|
|
# mno¾ina |
|
|
s/(${su})(ov|ev)*(i|a|ima|e|in|ina)$/$1\t6/g; |
|
|
# jednina |
|
|
s/(o|e|a|u|om|em)$/\t7/g; |
|
|
} |
|
81 |
|
|
82 |
|
# vrsta a |
83 |
|
s/me$/me 11/g || |
84 |
|
s/(eta|ena)$/e 12/g || |
85 |
|
|
86 |
|
s/(\w${sa}${su})(o|e|a|u|om|em|i|ima|ina|eta)$/$1 7/g || |
87 |
|
|
88 |
|
s/(${su})sa$/$1as 14/g || |
89 |
|
s/(${su})ena$/$1e 16/g || |
90 |
|
s/(${su})eta$/$1e 17/g || |
91 |
|
s/(${su})([oe])$/$1$2 18/g || |
92 |
|
|
93 |
printf("%-15s %s\n",$orig,$_); |
0; |
94 |
|
|
95 |
|
} |
96 |
|
|
97 |
|
s/\s\d+$//; |
98 |
} |
} |
99 |
|
|