1 |
#!/usr/bin/perl -w |
package StemHr; |
2 |
|
|
3 |
# |
# |
4 |
# Croatian stemmer |
# Croatian stemmer |
5 |
# |
# |
20 |
# |
# |
21 |
|
|
22 |
use strict; |
use strict; |
23 |
|
use locale; |
24 |
|
|
25 |
sub kgh { |
sub kgh { |
26 |
my ($pre,$replace,$post) = @_; |
my ($pre,$replace,$post) = @_; |
37 |
|
|
38 |
my $palatal = '(lj|nj|j|æ|¾|¹|¾|¹t|¾d)'; |
my $palatal = '(lj|nj|j|æ|¾|¹|¾|¹t|¾d)'; |
39 |
|
|
40 |
my %rules; |
# glagolni razredi |
41 |
my %stem_words; |
my $g_1r = '[td]'; |
42 |
my $words = 0; |
my $g_2r = '[sz]'; |
43 |
my $stems = 0; |
my $g_3r = '[pb]'; |
44 |
|
my $g_4r = '[kgh]'; |
45 |
|
my $g_5r = '[nm]'; |
46 |
my $last_stem = ''; |
my $g_6r = '(:?nu|n)'; |
|
sub check_stem { |
|
|
my $s = shift || return; |
|
|
if ($last_stem) { |
|
|
print "ERROR==> " if ($last_stem ne $s); |
|
|
} else { |
|
|
$last_stem = $s; |
|
|
} |
|
|
} |
|
|
|
|
|
while(<>) { |
|
|
chomp; |
|
|
next if (/^#/); |
|
|
if (/^$/) { |
|
|
print "\n"; |
|
|
$last_stem = ''; |
|
|
next; |
|
|
} |
|
47 |
|
|
48 |
$words++; |
sub stem { |
49 |
|
|
50 |
my $orig = $_; |
unless ( |
51 |
|
# infinitiv |
52 |
|
s/(\w)(ti|æi)$/$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)$/.æi 107/g || |
64 |
|
|
65 |
# imenice |
# imenice |
66 |
|
|
|
unless ( |
|
67 |
# vrsta a |
# vrsta a |
|
s/${palatal}e$/$1e 10/g || |
|
68 |
s/(${su}st)a$/$1 13/g || |
s/(${su}st)a$/$1 13/g || |
69 |
s/(${su})c[ae]$/$1ce 17/g || |
s/(${su})c[ae]$/$1ce 17/g || |
70 |
# kgh |
# kgh |
71 |
s/(${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,' 1')/gex || |
s/(\w${sa})([è¾¹czs])(i|e|ima)$/kgh($1,$2,' 1')/gex || |
72 |
s/(${sa}[kgh])(a|u|om)$/$1 2/g || |
s/(${sa}[kgh])(a|u|om)$/$1 2/g || |
73 |
s/(${su})([è¾¹czs])(i|e|ima)$/kgh($1,"a$2",' 3')/gex || |
s/(${su})([è¾¹czs])(i|e|ima)$/kgh($1,"a$2",' 3')/gex || |
74 |
# imenice na palatal |
# imenice na palatal |
82 |
# vrsta a |
# vrsta a |
83 |
s/me$/me 11/g || |
s/me$/me 11/g || |
84 |
s/(eta|ena)$/e 12/g || |
s/(eta|ena)$/e 12/g || |
85 |
s/(${sa}${su})(o|e|a|u|om|em|i|ima)$/$1 7/g; |
|
86 |
|
s/(\w${sa}${su})(o|e|a|u|om|em|i|ima|ina|eta)$/$1 7/g || |
87 |
|
|
88 |
s/(${su})sa$/$1as 14/g || |
s/(${su})sa$/$1as 14/g || |
89 |
s/(${su})ena$/$1e 16/g || |
s/(${su})ena$/$1e 16/g || |
90 |
s/eta$/e 17/g || |
s/(${su})eta$/$1e 17/g || |
91 |
s/(${su})([oe])$/$1-$2 18/g || |
s/(${su})([oe])$/$1$2 18/g || |
92 |
|
|
93 |
# mno¾ina |
0; |
|
s/(${su})(ov|ev)*(i|a|ima|e|in|ina|eta)$/$1 6/g; |
|
|
} |
|
94 |
|
|
|
if (s/^(.+)\s(\d+)$/$1\t$2/g) { |
|
|
$rules{$2}++; |
|
|
$stems++; |
|
|
$stem_words{$1}++; |
|
|
check_stem($1); |
|
|
} else { |
|
|
$last_stem = $_; |
|
95 |
} |
} |
96 |
|
|
97 |
printf("%-15s %s\n",$orig,$_); |
s/\s\d+$//; |
|
|
|
98 |
} |
} |
|
my $nr_stems = keys(%stem_words); |
|
|
printf "\n# %d words, %d stems in %d ops, %.2f%% size\n",$words,$nr_stems,$stems,($nr_stems*100/$words); |
|
99 |
|
|
|
foreach my $s (keys %stem_words) { |
|
|
print "#stem $stem_words{$s} $s\n"; |
|
|
} |
|
|
|
|
|
foreach my $r (sort keys %rules) { |
|
|
print "#rule $rules{$r} $r\n"; |
|
|
} |
|