1 |
dpavlin |
12 |
package StemHr; |
2 |
|
|
|
3 |
dpavlin |
1 |
# |
4 |
|
|
# Croatian stemmer |
5 |
|
|
# |
6 |
|
|
|
7 |
|
|
# promjenjive: |
8 |
|
|
# - imenice |
9 |
|
|
# - pridjevi |
10 |
|
|
# - brojevi |
11 |
|
|
# - zamjenice |
12 |
|
|
# - prilozi |
13 |
|
|
# - glagoli |
14 |
|
|
# |
15 |
|
|
# nepromjenjive: |
16 |
|
|
# - prijedlozi |
17 |
|
|
# - veznici |
18 |
|
|
# - èestice |
19 |
|
|
# - uzvici |
20 |
|
|
# |
21 |
|
|
|
22 |
dpavlin |
4 |
use strict; |
23 |
dpavlin |
12 |
use locale; |
24 |
dpavlin |
4 |
|
25 |
dpavlin |
1 |
sub kgh { |
26 |
|
|
my ($pre,$replace,$post) = @_; |
27 |
|
|
$replace =~ s/[cè]/k/g; |
28 |
|
|
$replace =~ s/[z¾]/g/g; |
29 |
|
|
$replace =~ s/[s¹]/h/g; |
30 |
dpavlin |
2 |
return $pre . $replace . $post; |
31 |
dpavlin |
1 |
} |
32 |
|
|
|
33 |
|
|
# samoglasnici |
34 |
|
|
my $sa = '[aeiou]'; |
35 |
|
|
# suglasnici |
36 |
|
|
my $su = '[^aeiou]'; |
37 |
|
|
|
38 |
dpavlin |
5 |
my $palatal = '(lj|nj|j|æ|¾|¹|¾|¹t|¾d)'; |
39 |
|
|
|
40 |
dpavlin |
7 |
# 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 $g_6r = '(:?nu|n)'; |
47 |
|
|
|
48 |
dpavlin |
12 |
sub stem { |
49 |
dpavlin |
3 |
|
50 |
dpavlin |
1 |
unless ( |
51 |
dpavlin |
7 |
# infinitiv |
52 |
dpavlin |
9 |
s/(\w)(ti|æi)$/$1.$2 100/g || |
53 |
dpavlin |
7 |
# 1. razred |
54 |
dpavlin |
9 |
s/([^sk])[td](em|e¹|e|emo|ete|oh|osmo|oste|o¹e|ijah|ija¹e|ijasno|ijaste|ijahu|imo|ite|en|ena|eni)$/$1s.ti 101/g || |
55 |
dpavlin |
7 |
# 2. razred |
56 |
dpavlin |
9 |
s/(${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)$/$1s.ti 102/ || |
57 |
dpavlin |
11 |
# 3. razred |
58 |
dpavlin |
10 |
s/(p|b|sp|su)(em|e¹|e|emo|ete|u|oh|osmo|oste|o¹e|ah|a¹e|asmo|aste|ahu|ijah|ija¹e|ijahu|i|imo|ite||uæi|av¹i|ao|la|lo|en|ena|eni)$/$1s.ti 103/ || |
59 |
dpavlin |
11 |
# 4. razred |
60 |
|
|
s/[è¾¹](em|e¹|e|emo|ete|u|ah|ahu|en|ena)$/.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)$/.æi 107/g || |
64 |
dpavlin |
7 |
|
65 |
|
|
# imenice |
66 |
|
|
|
67 |
dpavlin |
5 |
# vrsta a |
68 |
|
|
s/(${su}st)a$/$1 13/g || |
69 |
|
|
s/(${su})c[ae]$/$1ce 17/g || |
70 |
dpavlin |
2 |
# kgh |
71 |
dpavlin |
9 |
s/(\w${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,' 1')/gex || |
72 |
dpavlin |
2 |
s/(${sa}[kgh])(a|u|om)$/$1 2/g || |
73 |
|
|
s/(${su})([è¾¹czs])(i|e|ima)$/kgh($1,"a$2",' 3')/gex || |
74 |
dpavlin |
5 |
# imenice na palatal |
75 |
|
|
s/${palatal}${palatal}(a|u|em|i|ima|e)$/$1a$2 9/g || |
76 |
|
|
s/${palatal}(a|u|em)$/$1 8/g || |
77 |
dpavlin |
1 |
# nepostojano a |
78 |
dpavlin |
2 |
s/(${su}a{$su})a/$1 4/g || |
79 |
|
|
s/(${su})(${su})(a|u|i|e|om|ima)$/$1a$2 5/g |
80 |
dpavlin |
1 |
) { |
81 |
|
|
|
82 |
dpavlin |
5 |
# vrsta a |
83 |
|
|
s/me$/me 11/g || |
84 |
|
|
s/(eta|ena)$/e 12/g || |
85 |
|
|
|
86 |
dpavlin |
9 |
s/(\w${sa}${su})(o|e|a|u|om|em|i|ima|ina|eta)$/$1 7/g || |
87 |
|
|
|
88 |
dpavlin |
5 |
s/(${su})sa$/$1as 14/g || |
89 |
|
|
s/(${su})ena$/$1e 16/g || |
90 |
dpavlin |
6 |
s/(${su})eta$/$1e 17/g || |
91 |
dpavlin |
9 |
s/(${su})([oe])$/$1$2 18/g || |
92 |
dpavlin |
5 |
|
93 |
dpavlin |
9 |
0; |
94 |
|
|
|
95 |
dpavlin |
1 |
} |
96 |
|
|
|
97 |
dpavlin |
12 |
s/\s\d+$//; |
98 |
dpavlin |
1 |
} |
99 |
dpavlin |
3 |
|