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 |
use strict; |
22 |
|
23 |
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 |
return $pre . $replace . $post; |
29 |
} |
30 |
|
31 |
# samoglasnici |
32 |
my $sa = '[aeiou]'; |
33 |
# suglasnici |
34 |
my $su = '[^aeiou]'; |
35 |
|
36 |
my %rules; |
37 |
my %stem_words; |
38 |
my $words = 0; |
39 |
my $stems = 0; |
40 |
|
41 |
while(<>) { |
42 |
chomp; |
43 |
if (/^$/) { |
44 |
print "\n"; |
45 |
next; |
46 |
} |
47 |
|
48 |
$words++; |
49 |
|
50 |
my $orig = $_; |
51 |
|
52 |
# imenice |
53 |
# vrsta a |
54 |
|
55 |
unless ( |
56 |
# kgh |
57 |
s/(${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,' 1')/gex || |
58 |
s/(${sa}[kgh])(a|u|om)$/$1 2/g || |
59 |
s/(${su})([è¾¹czs])(i|e|ima)$/kgh($1,"a$2",' 3')/gex || |
60 |
# nepostojano a |
61 |
s/(${su}a{$su})a/$1 4/g || |
62 |
s/(${su})(${su})(a|u|i|e|om|ima)$/$1a$2 5/g |
63 |
) { |
64 |
|
65 |
# mno¾ina |
66 |
s/(${su})(ov|ev)*(i|a|ima|e|in|ina)$/$1 6/g; |
67 |
# jednina |
68 |
s/(o|e|a|u|om|em)$/ 7/g; |
69 |
} |
70 |
|
71 |
if (s/^(.+)\s(\d+)$/$1\t$2/g) { |
72 |
$rules{$2}++; |
73 |
$stems++; |
74 |
$stem_words{$1}++; |
75 |
} |
76 |
|
77 |
printf("%-15s %s\n",$orig,$_); |
78 |
|
79 |
} |
80 |
my $nr_stems = keys(%stem_words); |
81 |
printf "\n# %d words, %d stems in %d ops, %.2f%% size\n",$words,$nr_stems,$stems,($nr_stems*100/$words); |
82 |
|
83 |
foreach my $s (keys %stem_words) { |
84 |
print "#stem $stem_words{$s} $s\n"; |
85 |
} |
86 |
|
87 |
foreach my $r (keys %rules) { |
88 |
print "#rule $rules{$r} $r\n"; |
89 |
} |