Blame view
egs/lre07/v1/local/lre07_eval/lre07_targets.pl
3.34 KB
8dcb6dfcb
first commit
|
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 |
#!/usr/bin/env perl
#
# Copyright 2014 David Snyder
# Apache 2.0.
#
# Creates the target and nontarget files used by score_lre07.v01d.pl for
# NIST LRE 2007 General Language Recognition closed-set evaluation.
# See http://www.itl.nist.gov/iad/mig//tests/lre/2007/LRE07EvalPlan-v8b.pdf
# for more details on the evaluation.
if (@ARGV != 5) {
print STDERR "Usage: $0 <path-to-posteriors> <path-to-utt2lang> \
<path-to-languages.txt> <path-to-targets-output> \
<path-to-nontargets-output>
";
exit(1);
}
($posts, $utt2lang, $languages, $targets, $nontargets) = @ARGV;
%lang_to_idx = ();
%idx_to_lang = ();
%utt_to_lang = ();
$oos_lang = "zzz";
open(LANG2IDX, "<", $languages) || die "Cannot open $languages file";
while (<LANG2IDX>) {
chomp;
@toks = split(" ", $_);
$lang = $toks[0];
$idx = $toks[1];
$lang_to_idx{$lang} = $idx;
$idx_to_lang{$idx} = $lang;
}
close(LANG2IDX) || die;
open(UTT2LANG, "<", $utt2lang) || die "Cannot open $utt2lang file";
while (<UTT2LANG>) {
chomp;
@toks = split(" ", $_);
$utt = $toks[0];
$lang = $toks[1];
$utt_to_lang{$utt} = $lang;
}
close(UTT2LANG) || die;
open(POSTS, "<", $posts) || die "Cannot open $posts file";
open(TARGETS, ">", $targets) || die "Cannot open $targets file";
open(NONTARGETS, ">", $nontargets) || die "Cannot open $nontargets file";
while($line = <POSTS>) {
chomp($line);
$line =~ s/[\[\]]//g;
@toks = split(" ", $line);
$utt = $toks[0];
$actual_lang = $utt_to_lang{$utt};
$size = $#toks + 1;
$max_lang = "zzz";
$max_log_prob = -9**9**9; #-inf
$target_prob = 0;
# Handle target
for ($i = 1; $i < $size; $i++) {
if ($max_log_prob < $toks[$i]) {
$max_log_prob = $toks[$i];
$max_lang = $idx_to_lang{$i-1};
}
if ($actual_lang eq $idx_to_lang{$i-1}) {
print "$actual_lang $idx_to_lang{$i-1}
";
}
if (index($actual_lang, $idx_to_lang{$i-1}) != -1
|| $actual_lang eq $idx_to_lang{$i-1}) {
$target_prob = exp($toks[$i]);
}
}
if (index($actual_lang, ".") != -1) {
@lang_parts = split("[.]", $actual_lang);
$lang = $lang_parts[0];
} else {
$lang = $actual_lang;
}
if ($lang =~ /(arabic|bengali|farsi|german|japanese|korean|russian|tamil|thai|vietnamese|chinese|english|hindustani|spanish)/i) {
if (index($actual_lang, $max_lang) != -1 || $actual_lang eq $max_lang) {
print TARGETS "general_lr $lang closed_set $utt t $target_prob "
."$actual_lang
";
} else {
print TARGETS "general_lr $lang closed_set $utt f $target_prob "
."$actual_lang
";
}
}
# Handle nontarget
for ($i = 1; $i < $size; $i++) {
$nontarget_lang = $idx_to_lang{$i-1};
next if (index($actual_lang, $nontarget_lang) != -1
|| $actual_lang eq $nontarget_lang);
# if the nontarget lang is most probable
if ($nontarget_lang =~ /(arabic|bengali|farsi|german|japanese|korean|russian|tamil|thai|vietnamese|chinese|english|hindustani|spanish)/i) {
$prob = exp($toks[$i]);
if (index($max_lang, $nontarget_lang) != -1
|| $max_lang eq $nontarget_lang) {
print NONTARGETS "general_lr $nontarget_lang closed_set $utt t "
."$prob $actual_lang
";
} else {
print NONTARGETS "general_lr $nontarget_lang closed_set $utt f "
."$prob $actual_lang
";
}
}
}
}
close(POSTS) || die;
close(TARGETS) || die;
close(NONTARGETS) || die;
|