1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
|
#!/usr/bin/env perl
# Display OOV rate of a test set against a training corpus or a phrase table.
# Ondrej Bojar
use strict;
use warnings;
use Digest::MD5 qw(md5);
use Encode qw(encode_utf8);
use Getopt::Long;
binmode(STDIN, ":utf8");
binmode(STDOUT, ":utf8");
binmode(STDERR, ":utf8");
my $verbose = 0;
my $n = 1;
my $srcfile = undef;
my $md5 = 0;
GetOptions(
"n=i" => \$n, # the n-grams to search for (default: unigrams)
"verbose!" => \$verbose, # emit the list of oov words
"md5!" => \$md5, # emit the list of oov words
"src=s" => \$srcfile, # use this source file
) or exit 1;
my $testf = shift;
if (!defined $testf) {
print STDERR "usage: $0 test-corpus < training-corpus
Options:
--n=1 ... use phrases of n words as the unit
set --n=0 to compare *whole sentences* (forces md5 hashing on)
--md5 ... hash each ngram using md5, saves memory for longer n-grams
--verbose ... emit OOV phrases at the end
--src=test-src ... a word in the test-corpus not deemed OOV if present in the
corresponding source sentence in test-src.
The files test-corpus and test-src must be of equal length.
Synopsis:
Check OOV of a training corpus against a test set:
cat corpus.src.txt | $0 testset.txt
Check target-side OOV of a phrase table against a reference:
cat ttable | sed 's/ ||| /<ctrl-v><tab>/g' | cut -f 2 \
| $0 reference.txt
";
exit 1;
}
my $ngr_or_sent = $n > 0 ? "$n-grams" : "sentences";
# load source file to accept ngrams from source
my $source_confirms = undef;
my $srcfilelen = undef;
if (defined $srcfile) {
my $fh = my_open($srcfile);
my $nr = 0;
my $srctokens = 0;
while (<$fh>) {
$nr++;
chomp;
s/^\s+//;
s/\s+$//;
my $ngrams = ngrams($n, $_);
foreach my $ngr (keys %$ngrams) {
$source_confirms->[$nr]->{$ngr} += $ngrams->{$ngr};
$srctokens += $ngrams->{$ngr};
}
}
close $fh;
print "Source set sents\t$nr\n";
print "Source set running $ngr_or_sent\t$srctokens\n" if $n>0;
$srcfilelen = $nr;
}
my %needed = ();
my $fh = my_open($testf);
my $nr = 0;
my $testtokens = 0;
my %testtypes = ();
while (<$fh>) {
$nr++;
chomp;
s/^\s+//;
s/\s+$//;
my $ngrams = ngrams($n, $_);
foreach my $ngr (keys %$ngrams) {
$needed{$ngr} += $ngrams->{$ngr}
unless $source_confirms->[$nr]->{$ngr};
$testtokens += $ngrams->{$ngr};
$testtypes{$ngr} = 1;
}
}
close $fh;
my $testtypesneeded = scalar(keys(%needed));
my $testtypes = scalar(keys(%testtypes));
print "Test set sents\t$nr\n";
print "Test set running $n-grams\t$testtokens\n" if $n>0;
print "Test set unique $ngr_or_sent needed\t$testtypesneeded\n";
print "Test set unique $ngr_or_sent\t$testtypes\n";
die "Mismatching sent count: $srcfile and $testf ($srcfilelen vs. $nr)"
if defined $srcfile && $srcfilelen != $nr;
my %seen;
$nr = 0;
my $traintokens = 0;
while (<>) {
$nr++;
print STDERR "." if $nr % 10000 == 0;
print STDERR "($nr)" if $nr % 500000 == 0;
chomp;
s/^\s+//;
s/\s+$//;
my $ngrams = ngrams($n, $_); # [ split /\s+/, $_ ]);
foreach my $ngr (keys %$ngrams) {
$seen{$ngr} = 1 if $ngrams->{$ngr};
$traintokens += $ngrams->{$ngr};
}
}
foreach my $ngr (keys %needed) {
delete $needed{$ngr} if defined $seen{$ngr};
}
print STDERR "Done.\n";
my $traintypes = scalar(keys(%seen));
print "Training set sents\t$nr\n";
print "Training set running $n-grams\t$traintokens\n" if $n>0;
print "Training set unique $ngr_or_sent\t$traintypes\n";
my $oovtypes = scalar(keys(%needed));
my $oovtokens = 0;
foreach my $v (values %needed) {
$oovtokens += $v;
}
printf "OOV $ngr_or_sent types\t%i\t%.1f %%\n", $oovtypes, $oovtypes/$testtypes*100;
printf "OOV $ngr_or_sent tokens\t%i\t%.1f %%\n", $oovtokens, $oovtokens/$testtokens*100;
if ($verbose) {
foreach my $ngr (sort {$needed{$b} <=> $needed{$a}} keys %needed) {
print "$needed{$ngr}\t$ngr\n";
}
}
sub my_open {
my $f = shift;
if ($f eq "-") {
binmode(STDIN, ":utf8");
return *STDIN;
}
die "Not found: $f" if ! -e $f;
my $opn;
my $hdl;
my $ft = `file '$f'`;
# file might not recognize some files!
if ($f =~ /\.gz$/ || $ft =~ /gzip compressed data/) {
$opn = "zcat '$f' |";
} elsif ($f =~ /\.bz2$/ || $ft =~ /bzip2 compressed data/) {
$opn = "bzcat '$f' |";
} else {
$opn = "$f";
}
open $hdl, $opn or die "Can't open '$opn': $!";
binmode $hdl, ":utf8";
return $hdl;
}
sub ngrams {
my $n = shift;
my $sent = shift;
if ($n == 0) {
return { md5(encode_utf8($sent)) => 1 };
} else {
my @words = split /\s+/, $sent;
my $out;
if ($n == 1) {
foreach my $w (@words) {
my $usew = $md5 ? md5(encode_utf8($$w)) : $w;
$out->{$w}++;
}
} else {
while ($#words >= $n-1) {
my $ngr = join(" ", @words[0..$n-1]);
my $usengr = $md5 ? md5(encode_utf8($ngr)) : $ngr;
$out->{$ngr}++;
shift @words;
}
}
return $out;
}
}
|
