1 |
dpavlin |
1 |
#!/usr/bin/perl -w |
2 |
|
|
# |
3 |
|
|
# Croatian stemmer |
4 |
|
|
# |
5 |
|
|
|
6 |
|
|
# promjenjive: |
7 |
|
|
# - imenice |
8 |
|
|
# - pridjevi |
9 |
|
|
# - brojevi |
10 |
|
|
# - zamjenice |
11 |
|
|
# - prilozi |
12 |
|
|
# - glagoli |
13 |
|
|
# |
14 |
|
|
# nepromjenjive: |
15 |
|
|
# - prijedlozi |
16 |
|
|
# - veznici |
17 |
|
|
# - èestice |
18 |
|
|
# - uzvici |
19 |
|
|
# |
20 |
|
|
|
21 |
dpavlin |
4 |
use strict; |
22 |
|
|
|
23 |
dpavlin |
1 |
sub kgh { |
24 |
|
|
my ($pre,$replace,$post) = @_; |
25 |
|
|
$replace =~ s/[cè]/k/g; |
26 |
|
|
$replace =~ s/[z¾]/g/g; |
27 |
|
|
$replace =~ s/[s¹]/h/g; |
28 |
dpavlin |
2 |
return $pre . $replace . $post; |
29 |
dpavlin |
1 |
} |
30 |
|
|
|
31 |
|
|
# samoglasnici |
32 |
|
|
my $sa = '[aeiou]'; |
33 |
|
|
# suglasnici |
34 |
|
|
my $su = '[^aeiou]'; |
35 |
|
|
|
36 |
dpavlin |
5 |
my $palatal = '(lj|nj|j|æ|¾|¹|¾|¹t|¾d)'; |
37 |
|
|
|
38 |
dpavlin |
7 |
# glagolni razredi |
39 |
|
|
my $g_1r = '[td]'; |
40 |
|
|
my $g_2r = '[sz]'; |
41 |
|
|
my $g_3r = '[pb]'; |
42 |
|
|
my $g_4r = '[kgh]'; |
43 |
|
|
my $g_5r = '[nm]'; |
44 |
|
|
my $g_6r = '(:?nu|n)'; |
45 |
|
|
|
46 |
dpavlin |
3 |
my %rules; |
47 |
|
|
my %stem_words; |
48 |
|
|
my $words = 0; |
49 |
|
|
my $stems = 0; |
50 |
|
|
|
51 |
dpavlin |
5 |
|
52 |
|
|
my $last_stem = ''; |
53 |
|
|
sub check_stem { |
54 |
|
|
my $s = shift || return; |
55 |
|
|
if ($last_stem) { |
56 |
|
|
print "ERROR==> " if ($last_stem ne $s); |
57 |
|
|
} else { |
58 |
|
|
$last_stem = $s; |
59 |
|
|
} |
60 |
|
|
} |
61 |
|
|
|
62 |
dpavlin |
1 |
while(<>) { |
63 |
|
|
chomp; |
64 |
dpavlin |
5 |
next if (/^#/); |
65 |
dpavlin |
3 |
if (/^$/) { |
66 |
|
|
print "\n"; |
67 |
dpavlin |
5 |
$last_stem = ''; |
68 |
dpavlin |
3 |
next; |
69 |
|
|
} |
70 |
dpavlin |
1 |
|
71 |
dpavlin |
3 |
$words++; |
72 |
|
|
|
73 |
dpavlin |
1 |
my $orig = $_; |
74 |
|
|
|
75 |
|
|
|
76 |
|
|
unless ( |
77 |
dpavlin |
7 |
# infinitiv |
78 |
|
|
s/(${sa})(ti|æi)$/$1.$2 100/g || |
79 |
|
|
# 1. razred |
80 |
|
|
s/([^s])[td](em|e¹|emo|ete|oh|osmo|oste|o¹e|ijah|ija¹e|ijasno|ijaste|ijahu|imo|ite|en|ena|eni)$/$1.sti 101/g || |
81 |
|
|
# 2. razred |
82 |
|
|
s/(${sa}[sz])(em|e¹|e¹emo|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)$/$1.sti 102/ || |
83 |
|
|
# 3. razdred |
84 |
|
|
s/(p|b|sp|su)(em|e¹|e|emo|ete|u|oh|osmo|oste|o¹e|ah|a¹e|asmo|aste|ahu|ijah|ija¹e|i|imo|ite||uæi|av¹i|ao|la|lo|en|ena|eni)$/$1.sti 103/ || |
85 |
|
|
|
86 |
|
|
# imenice |
87 |
|
|
|
88 |
dpavlin |
5 |
# vrsta a |
89 |
|
|
s/(${su}st)a$/$1 13/g || |
90 |
|
|
s/(${su})c[ae]$/$1ce 17/g || |
91 |
dpavlin |
2 |
# kgh |
92 |
|
|
s/(${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,' 1')/gex || |
93 |
|
|
s/(${sa}[kgh])(a|u|om)$/$1 2/g || |
94 |
|
|
s/(${su})([è¾¹czs])(i|e|ima)$/kgh($1,"a$2",' 3')/gex || |
95 |
dpavlin |
5 |
# imenice na palatal |
96 |
|
|
s/${palatal}${palatal}(a|u|em|i|ima|e)$/$1a$2 9/g || |
97 |
|
|
s/${palatal}(a|u|em)$/$1 8/g || |
98 |
dpavlin |
1 |
# nepostojano a |
99 |
dpavlin |
2 |
s/(${su}a{$su})a/$1 4/g || |
100 |
|
|
s/(${su})(${su})(a|u|i|e|om|ima)$/$1a$2 5/g |
101 |
dpavlin |
1 |
) { |
102 |
|
|
|
103 |
dpavlin |
5 |
# vrsta a |
104 |
|
|
s/me$/me 11/g || |
105 |
|
|
s/(eta|ena)$/e 12/g || |
106 |
dpavlin |
6 |
s/(${sa}${su})(o|e|a|u|om|em|i|ima|ina)$/$1 7/g || |
107 |
dpavlin |
5 |
|
108 |
|
|
s/(${su})sa$/$1as 14/g || |
109 |
|
|
s/(${su})ena$/$1e 16/g || |
110 |
dpavlin |
6 |
s/(${su})eta$/$1e 17/g || |
111 |
dpavlin |
5 |
s/(${su})([oe])$/$1-$2 18/g || |
112 |
|
|
|
113 |
dpavlin |
1 |
# mno¾ina |
114 |
dpavlin |
5 |
s/(${su})(ov|ev)*(i|a|ima|e|in|ina|eta)$/$1 6/g; |
115 |
dpavlin |
1 |
} |
116 |
|
|
|
117 |
dpavlin |
3 |
if (s/^(.+)\s(\d+)$/$1\t$2/g) { |
118 |
|
|
$rules{$2}++; |
119 |
|
|
$stems++; |
120 |
|
|
$stem_words{$1}++; |
121 |
dpavlin |
5 |
check_stem($1); |
122 |
|
|
} else { |
123 |
|
|
$last_stem = $_; |
124 |
dpavlin |
3 |
} |
125 |
dpavlin |
1 |
|
126 |
|
|
printf("%-15s %s\n",$orig,$_); |
127 |
dpavlin |
3 |
|
128 |
dpavlin |
1 |
} |
129 |
dpavlin |
3 |
my $nr_stems = keys(%stem_words); |
130 |
|
|
printf "\n# %d words, %d stems in %d ops, %.2f%% size\n",$words,$nr_stems,$stems,($nr_stems*100/$words); |
131 |
|
|
|
132 |
dpavlin |
4 |
foreach my $s (keys %stem_words) { |
133 |
|
|
print "#stem $stem_words{$s} $s\n"; |
134 |
|
|
} |
135 |
|
|
|
136 |
dpavlin |
5 |
foreach my $r (sort keys %rules) { |
137 |
dpavlin |
4 |
print "#rule $rules{$r} $r\n"; |
138 |
|
|
} |