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 |
|
|
sub kgh { |
22 |
|
|
my ($pre,$replace,$post) = @_; |
23 |
|
|
$replace =~ s/[cè]/k/g; |
24 |
|
|
$replace =~ s/[z¾]/g/g; |
25 |
|
|
$replace =~ s/[s¹]/h/g; |
26 |
dpavlin |
2 |
return $pre . $replace . $post; |
27 |
dpavlin |
1 |
} |
28 |
|
|
|
29 |
|
|
# samoglasnici |
30 |
|
|
my $sa = '[aeiou]'; |
31 |
|
|
# suglasnici |
32 |
|
|
my $su = '[^aeiou]'; |
33 |
|
|
|
34 |
dpavlin |
3 |
my %rules; |
35 |
|
|
my %stem_words; |
36 |
|
|
my $words = 0; |
37 |
|
|
my $stems = 0; |
38 |
|
|
|
39 |
dpavlin |
1 |
while(<>) { |
40 |
|
|
chomp; |
41 |
dpavlin |
3 |
if (/^$/) { |
42 |
|
|
print "\n"; |
43 |
|
|
next; |
44 |
|
|
} |
45 |
dpavlin |
1 |
|
46 |
dpavlin |
3 |
$words++; |
47 |
|
|
|
48 |
dpavlin |
1 |
my $orig = $_; |
49 |
|
|
|
50 |
|
|
# imenice |
51 |
|
|
# vrsta a |
52 |
|
|
|
53 |
|
|
unless ( |
54 |
dpavlin |
2 |
# kgh |
55 |
|
|
s/(${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,' 1')/gex || |
56 |
|
|
s/(${sa}[kgh])(a|u|om)$/$1 2/g || |
57 |
|
|
s/(${su})([è¾¹czs])(i|e|ima)$/kgh($1,"a$2",' 3')/gex || |
58 |
dpavlin |
1 |
# nepostojano a |
59 |
dpavlin |
2 |
s/(${su}a{$su})a/$1 4/g || |
60 |
|
|
s/(${su})(${su})(a|u|i|e|om|ima)$/$1a$2 5/g |
61 |
dpavlin |
1 |
) { |
62 |
|
|
|
63 |
|
|
# mno¾ina |
64 |
dpavlin |
2 |
s/(${su})(ov|ev)*(i|a|ima|e|in|ina)$/$1 6/g; |
65 |
dpavlin |
1 |
# jednina |
66 |
dpavlin |
2 |
s/(o|e|a|u|om|em)$/ 7/g; |
67 |
dpavlin |
1 |
} |
68 |
|
|
|
69 |
dpavlin |
3 |
if (s/^(.+)\s(\d+)$/$1\t$2/g) { |
70 |
|
|
$rules{$2}++; |
71 |
|
|
$stems++; |
72 |
|
|
$stem_words{$1}++; |
73 |
|
|
} |
74 |
dpavlin |
1 |
|
75 |
|
|
printf("%-15s %s\n",$orig,$_); |
76 |
dpavlin |
3 |
|
77 |
dpavlin |
1 |
} |
78 |
dpavlin |
3 |
my $nr_stems = keys(%stem_words); |
79 |
|
|
printf "\n# %d words, %d stems in %d ops, %.2f%% size\n",$words,$nr_stems,$stems,($nr_stems*100/$words); |
80 |
|
|
|