Blame view
egs/wsj/s5/steps/cleanup/create_segments_from_ctm.pl
18.2 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 |
#!/usr/bin/env perl
# Copyright 2014 Guoguo Chen; 2015 Nagendra Kumar Goel
# Apache 2.0
use strict;
use warnings;
use Getopt::Long;
# $SIG{__WARN__} = sub { $DB::single = 1 };
my $Usage = <<'EOU';
This script creates the segments file and text file for a data directory with
new segmentation. It takes a ctm file and an "alignment" file. The ctm file
corresponds to the audio that we want to make segmentations for, and is created
by decoding the audio using existing in-domain models. The "alignment" file is
generated by the binary align-text, and is Levenshtein alignment between the
original transcript and the decoded output.
Internally, the script first tries to find silence regions (gaps in the CTM).
If a silence region is found, and the neighboring words are free of errors
according to the alignment file, then this silence region will be taken as
a split point, and new segment will be created. If the new segment we are going
to output is too long (longer than --max-seg-length), the script will split
the long segments into smaller pieces with length roughly --max-seg-length.
If you are going to use --wer-cutoff to filter out segments with high WER, make
sure you set it to a reasonable value. If the value you set is higher than the
WER from your alignment file, then most of the segments will be filtered out.
Usage: steps/cleanup/create_segments_from_ctm.pl [options] \
<ctm> <aligned.txt> <segments> <text>
e.g.: steps/cleanup/create_segments_from_ctm.pl \
train_si284_split.ctm train_si284_split.aligned.txt \
data/train_si284_reseg/segments data/train_si284_reseg/text
Allowed options:
--max-seg-length : Maximum length of new segments (default = 10.0)
--min-seg-length : Minimum length of new segments (default = 2.0)
--min-sil-length : Minimum length of silence as split point (default = 0.5)
--separator : Separator for aligned pairs (default = ";")
--special-symbol : Special symbol to aligned with inserted or deleted words
(default = "<***>")
--wer-cutoff : Ignore segments with WER higher than the specified value.
-1 means no segment will be ignored. (default = -1)
--use-silence-midpoints : Set to 1 if you want to use silence midpoints
instead of min_sil_length for silence overhang.(default 0)
--force-correct-boundary-words : Set to zero if the segments will not be
required to have boundary words to be correct. Default 1
--aligned-ctm-filename : If set, the intermediate aligned ctm
is saved to this file
EOU
my $max_seg_length = 10.0;
my $min_seg_length = 2.0;
my $min_sil_length = 0.5;
my $separator = ";";
my $special_symbol = "<***>";
my $wer_cutoff = -1;
my $use_silence_midpoints = 0;
my $force_correct_boundary_words = 1;
my $aligned_ctm_filename = "";
GetOptions(
'wer-cutoff=f' => \$wer_cutoff,
'max-seg-length=f' => \$max_seg_length,
'min-seg-length=f' => \$min_seg_length,
'min-sil-length=f' => \$min_sil_length,
'use-silence-midpoints=f' => \$use_silence_midpoints,
'force-correct-boundary-words=f' => \$force_correct_boundary_words,
'aligned-ctm-filename=s' => \$aligned_ctm_filename,
'separator=s' => \$separator,
'special-symbol=s' => \$special_symbol);
if (@ARGV != 4) {
die $Usage;
}
my ($ctm_in, $align_in, $segments_out, $text_out) = @ARGV;
open(CI, "<$ctm_in") || die "Error: fail to open $ctm_in
";
open(AI, "<$align_in") || die "Error: fail to open $align_in
";
open(my $SO, ">$segments_out") || die "Error: fail to open $segments_out
";
open(my $TO, ">$text_out") || die "Error: fail to open $text_out
";
my $ACT= undef;
if ($aligned_ctm_filename ne "") {
open($ACT, ">$aligned_ctm_filename");
}
# Prints the current segment to file.
sub PrintSegment {
my ($aligned_ctm, $wav_id, $min_sil_length, $min_seg_length,
$seg_start_index, $seg_end_index, $seg_count, $SO, $TO) = @_;
if ($seg_start_index > $seg_end_index) {
return -1;
}
# Removes the surrounding silence.
while ($seg_start_index < scalar(@{$aligned_ctm}) &&
$aligned_ctm->[$seg_start_index]->[0] eq "<eps>") {
$seg_start_index += 1;
}
while ($seg_end_index >= 0 &&
$aligned_ctm->[$seg_end_index]->[0] eq "<eps>") {
$seg_end_index -= 1;
}
if ($seg_start_index > $seg_end_index) {
return -1;
}
# Filters out segments with high WER.
if ($wer_cutoff != -1) {
my $num_errors = 0; my $num_words = 0;
for (my $i = $seg_start_index; $i <= $seg_end_index; $i += 1) {
if ($aligned_ctm->[$i]->[0] ne "<eps>") {
$num_words += 1;
}
$num_errors += $aligned_ctm->[$i]->[3];
}
if ($num_errors / $num_words > $wer_cutoff || $num_words < 1) {
return -1;
}
}
# Works out the surrounding silence.
my $index = $seg_start_index - 1;
while ($index >= 0 && $aligned_ctm->[$index]->[0] eq
"<eps>" && $aligned_ctm->[$index]->[3] == 0) {
$index -= 1;
}
my $left_of_segment_has_deletion = "false";
$left_of_segment_has_deletion = "true"
if ($index > 0 && $aligned_ctm->[$index-1]->[0] ne "<eps>"
&& $aligned_ctm->[$index-1]->[3] == 0);
my $pad_start_sil = ($aligned_ctm->[$seg_start_index]->[1] -
$aligned_ctm->[$index + 1]->[1]) / 2.0;
if (($left_of_segment_has_deletion eq "true") || !$use_silence_midpoints) {
if ($pad_start_sil > $min_sil_length / 2.0) {
$pad_start_sil = $min_sil_length / 2.0;
}
}
my $right_of_segment_has_deletion = "false";
$index = $seg_end_index + 1;
while ($index < scalar(@{$aligned_ctm}) &&
$aligned_ctm->[$index]->[0] eq "<eps>" &&
$aligned_ctm->[$index]->[3] == 0) {
$index += 1;
}
$right_of_segment_has_deletion = "true"
if ($index < scalar(@{$aligned_ctm})-1 && $aligned_ctm->[$index+1]->[0] ne
"<eps>" && $aligned_ctm->[$index - 1]->[3] > 0);
my $pad_end_sil = ($aligned_ctm->[$index - 1]->[1] +
$aligned_ctm->[$index - 1]->[2] -
$aligned_ctm->[$seg_end_index]->[1] -
$aligned_ctm->[$seg_end_index]->[2]) / 2.0;
if (($right_of_segment_has_deletion eq "true") || !$use_silence_midpoints) {
if ($pad_end_sil > $min_sil_length / 2.0) {
$pad_end_sil = $min_sil_length / 2.0;
}
}
my $seg_start = $aligned_ctm->[$seg_start_index]->[1] - $pad_start_sil;
my $seg_end = $aligned_ctm->[$seg_end_index]->[1] +
$aligned_ctm->[$seg_end_index]->[2] + $pad_end_sil;
if ($seg_end - $seg_start < $min_seg_length) {
return -1;
}
$seg_start = sprintf("%.2f", $seg_start);
$seg_end = sprintf("%.2f", $seg_end);
my $seg_id = $wav_id . "_" . sprintf("%05d", $seg_count);
print $SO "$seg_id $wav_id $seg_start $seg_end
";
print $TO "$seg_id ";
for (my $x = $seg_start_index; $x <= $seg_end_index; $x += 1) {
if ($aligned_ctm->[$x]->[0] ne "<eps>") {
print $TO "$aligned_ctm->[$x]->[0] ";
}
}
print $TO "
";
return 0;
}
# Computes split point.
sub GetSplitPoint {
my ($aligned_ctm, $seg_start_index, $seg_end_index, $max_seg_length) = @_;
# Scan in the reversed order so we can maximize the length.
my $split_point = $seg_start_index;
for (my $x = $seg_end_index; $x > $seg_start_index; $x -= 1) {
my $current_seg_length = $aligned_ctm->[$x]->[1] +
$aligned_ctm->[$x]->[2] -
$aligned_ctm->[$seg_start_index]->[1];
if ($current_seg_length <= $max_seg_length) {
$split_point = $x;
last;
}
}
return $split_point;
}
# Computes segment length without surrounding silence.
sub GetSegmentLengthNoSil {
my ($aligned_ctm, $seg_start_index, $seg_end_index) = @_;
while ($seg_start_index < scalar(@{$aligned_ctm}) &&
$aligned_ctm->[$seg_start_index]->[0] eq "<eps>") {
$seg_start_index += 1;
}
while ($seg_end_index >= 0 &&
$aligned_ctm->[$seg_end_index]->[0] eq "<eps>") {
$seg_end_index -= 1;
}
if ($seg_start_index > $seg_end_index) {
return 0;
}
my $current_seg_length = $aligned_ctm->[$seg_end_index]->[1] +
$aligned_ctm->[$seg_end_index]->[2] -
$aligned_ctm->[$seg_start_index]->[1];
return $current_seg_length;
}
# Force splits long segments.
sub SplitLongSegment {
my ($aligned_ctm, $wav_id, $max_seg_length, $min_sil_length,
$seg_start_index, $seg_end_index, $current_seg_count, $SO, $TO) = @_;
# If the segment is too long, we manually split it. We make sure that the
# resulting segments are at least ($max_seg_length / 2) seconds long.
my $current_seg_length = $aligned_ctm->[$seg_end_index]->[1] +
$aligned_ctm->[$seg_end_index]->[2] -
$aligned_ctm->[$seg_start_index]->[1];
my $current_seg_index = $seg_start_index;
my $aligned_ctm_size = scalar(@{$aligned_ctm});
while ($current_seg_length > 1.5 * $max_seg_length && $current_seg_index < $aligned_ctm_size-1) {
my $split_point = GetSplitPoint($aligned_ctm, $current_seg_index,
$seg_end_index, $max_seg_length);
my $ans = PrintSegment($aligned_ctm, $wav_id, $min_sil_length,
$min_seg_length, $current_seg_index, $split_point,
$current_seg_count, $SO, $TO);
$current_seg_count += 1 if ($ans != -1);
$current_seg_index = $split_point + 1;
$current_seg_length = $aligned_ctm->[$seg_end_index]->[1] +
$aligned_ctm->[$seg_end_index]->[2] -
$aligned_ctm->[$current_seg_index]->[1];
}
if ($current_seg_index eq $aligned_ctm_size-1) {
my $ans = PrintSegment($aligned_ctm, $wav_id, $min_sil_length,
$min_seg_length, $current_seg_index, $current_seg_index,
$current_seg_count, $SO, $TO);
$current_seg_count += 1 if ($ans != -1);
return ($current_seg_count, $current_seg_index);
}
if ($current_seg_length > $max_seg_length) {
my $split_point = GetSplitPoint($aligned_ctm, $current_seg_index,
$seg_end_index,
$current_seg_length / 2.0 + 0.01);
my $ans = PrintSegment($aligned_ctm, $wav_id, $min_sil_length,
$min_seg_length, $current_seg_index, $split_point,
$current_seg_count, $SO, $TO);
$current_seg_count += 1 if ($ans != -1);
$current_seg_index = $split_point + 1;
}
my $split_point = GetSplitPoint($aligned_ctm, $current_seg_index,
$seg_end_index, $max_seg_length + 0.01);
my $ans = PrintSegment($aligned_ctm, $wav_id, $min_sil_length,
$min_seg_length, $current_seg_index, $split_point,
$current_seg_count, $SO, $TO);
$current_seg_count += 1 if ($ans != -1);
$current_seg_index = $split_point + 1;
return ($current_seg_count, $current_seg_index);
}
# Processes each wav file.
sub ProcessWav {
my ($max_seg_length, $min_seg_length, $min_sil_length, $special_symbol,
$current_ctm, $current_align, $SO, $TO, $ACT) = @_;
my $wav_id = $current_ctm->[0]->[0];
my $channel_id = $current_ctm->[0]->[1];
defined($wav_id) || die "Error: empty wav section
";
# First, we have to align the ctm file to the Levenshtein alignment.
# @aligned_ctm is a list of the following:
# [word, start_time, duration, num_errors]
my $ctm_index = 0;
my @aligned_ctm = ();
foreach my $entry (@{$current_align}) {
my $ref_word = $entry->[0];
my $hyp_word = $entry->[1];
if ($hyp_word eq $special_symbol) {
# Case 1: deletion, $hyp does not correspond to a word in the ctm file.
my $start = 0.0; my $dur = 0.0;
if (defined($aligned_ctm[-1])) {
$start = $aligned_ctm[-1]->[1] + $aligned_ctm[-1]->[2];
}
push(@aligned_ctm, [$ref_word, $start, $dur, 1]);
} else {
# Case 2: non-deletion, now $hyp corresponds to a word in ctm file.
while ($current_ctm->[$ctm_index]->[4] eq "<eps>") {
# Case 2.1: ctm contains silence at the corresponding place.
push(@aligned_ctm, ["<eps>", $current_ctm->[$ctm_index]->[2],
$current_ctm->[$ctm_index]->[3], 0]);
$ctm_index += 1;
}
my $ctm_word = $current_ctm->[$ctm_index]->[4];
$hyp_word eq $ctm_word ||
die "Error: got word $hyp_word in alignment but $ctm_word in ctm
";
my $start = $current_ctm->[$ctm_index]->[2];
my $dur = $current_ctm->[$ctm_index]->[3];
if ($ref_word ne $ctm_word) {
if ($ref_word eq $special_symbol) {
# Case 2.2: insertion, we propagate the duration and error to the
# previous one.
if (defined($aligned_ctm[-1])) {
$aligned_ctm[-1]->[2] += $dur;
$aligned_ctm[-1]->[3] += 1;
} else {
push(@aligned_ctm, ["<eps>", $start, $dur, 1]);
}
} else {
# Case 2.3: substitution.
push(@aligned_ctm, [$ref_word, $start, $dur, 1]);
}
} else {
# Case 2.4: correct.
push(@aligned_ctm, [$ref_word, $start, $dur, 0]);
}
$ctm_index += 1;
}
}
# Save the aligned CTM if needed
if(defined($ACT)){
for (my $i = 0; $i <= $#aligned_ctm; $i++) {
print $ACT "$wav_id $channel_id $aligned_ctm[$i][1] $aligned_ctm[$i][2] ";
print $ACT "$aligned_ctm[$i][0] $aligned_ctm[$i][3]
";
}
}
# Second, we create segments from @align_ctm, using simple greedy method.
my $current_seg_index = 0;
my $current_seg_count = 0;
for (my $x = 0; $x < @aligned_ctm; $x += 1) {
my $lcorrect = "true"; my $rcorrect = "true";
$lcorrect = "false" if ($x > 0 && $aligned_ctm[$x - 1]->[3] > 0);
$rcorrect = "false" if ($x < @aligned_ctm - 1 &&
$aligned_ctm[$x + 1]->[3] > 0);
my $current_seg_length = GetSegmentLengthNoSil(\@aligned_ctm,
$current_seg_index, $x);
# We split the audio, if the silence is longer than the requested silence
# length, and if there are no alignment error around it. We also make sure
# that segment contains actual words, instead of pure silence.
if ($aligned_ctm[$x]->[0] eq "<eps>" &&
$aligned_ctm[$x]->[2] >= $min_sil_length
&& (($force_correct_boundary_words && $lcorrect eq "true" &&
$rcorrect eq "true") || !$force_correct_boundary_words)) {
if ($current_seg_length <= $max_seg_length &&
$current_seg_length >= $min_seg_length) {
my $ans = PrintSegment(\@aligned_ctm, $wav_id, $min_sil_length,
$min_seg_length, $current_seg_index, $x,
$current_seg_count, $SO, $TO);
$current_seg_count += 1 if ($ans != -1);
$current_seg_index = $x + 1;
} elsif ($current_seg_length > $max_seg_length) {
($current_seg_count, $current_seg_index)
= SplitLongSegment(\@aligned_ctm, $wav_id, $max_seg_length,
$min_sil_length, $current_seg_index, $x,
$current_seg_count, $SO, $TO);
}
}
}
# Last segment.
if ($current_seg_index <= @aligned_ctm - 1) {
SplitLongSegment(\@aligned_ctm, $wav_id, $max_seg_length, $min_sil_length,
$current_seg_index, @aligned_ctm - 1,
$current_seg_count, $SO, $TO);
}
}
# Insert <eps> as silence so the down stream process will be easier. Example:
#
# Input ctm:
# 011 A 3.39 0.23 SELL
# 011 A 3.62 0.18 OFF
# 011 A 3.83 0.45 ASSETS
#
# Output ctm:
# 011 A 3.39 0.23 SELL
# 011 A 3.62 0.18 OFF
# 011 A 3.80 0.03 <eps>
# 011 A 3.83 0.45 ASSETS
sub InsertSilence {
my ($ctm_in, $ctm_out) = @_;
for (my $x = 1; $x < @{$ctm_in}; $x += 1) {
push(@{$ctm_out}, $ctm_in->[$x - 1]);
my $new_start = sprintf("%.2f",
$ctm_in->[$x - 1]->[2] + $ctm_in->[$x - 1]->[3]);
if ($new_start < $ctm_in->[$x]->[2]) {
my $new_dur = sprintf("%.2f", $ctm_in->[$x]->[2] - $new_start);
push(@{$ctm_out}, [$ctm_in->[$x - 1]->[0], $ctm_in->[$x - 1]->[1],
$new_start, $new_dur, "<eps>"]);
}
}
push(@{$ctm_out}, $ctm_in->[@{$ctm_in} - 1]);
}
# Reads the alignment.
my %aligned = ();
while (<AI>) {
chomp;
my @col = split;
@col >= 2 || die "Error: bad line $_
";
my $wav = shift @col;
if ( (@col + 0) % 3 != 2) {
die "Bad line in align-text output (unexpected number of fields): $_";
}
my @pairs = ();
for (my $x = 0; $x * 3 + 2 < @col; $x++) {
my $first_word = $col[$x * 3];
my $second_word = $col[$x * 3 + 1];
if ($x * 3 + 2 < @col) {
if ($col[$x * 3 + 2] ne $separator) {
die "Bad line in align-text output (expected separator '$separator'): $_";
}
}
# the [ ] expression returns a reference to a new anonymous array.
push(@pairs, [ $first_word, $second_word ]);
}
! defined($aligned{$wav}) || die "Error: $wav has already been processed
";
$aligned{$wav} = \@pairs;
}
# Reads the ctm file and creates the segmentation.
my $previous_wav_id = "";
my $previous_channel_id = "";
my @current_wav = ();
while (<CI>) {
chomp;
my @col = split;
@col >= 5 || die "Error: bad line $_
";
if ($previous_wav_id eq $col[0] && $previous_channel_id eq $col[1]) {
push(@current_wav, \@col);
} else {
if (@current_wav > 0) {
defined($aligned{$previous_wav_id}) ||
die "Error: no alignment info for $previous_wav_id
";
my @current_wav_silence = ();
InsertSilence(\@current_wav, \@current_wav_silence);
ProcessWav($max_seg_length, $min_seg_length, $min_sil_length,
$special_symbol, \@current_wav_silence,
$aligned{$previous_wav_id}, $SO, $TO, $ACT);
}
@current_wav = ();
push(@current_wav, \@col);
$previous_wav_id = $col[0];
$previous_channel_id = $col[1];
}
}
# The last wav file.
if (@current_wav > 0) {
defined($aligned{$previous_wav_id}) ||
die "Error: no alignment info for $previous_wav_id
";
my @current_wav_silence = ();
InsertSilence(\@current_wav, \@current_wav_silence);
ProcessWav($max_seg_length, $min_seg_length, $min_sil_length, $special_symbol,
\@current_wav_silence, $aligned{$previous_wav_id}, $SO, $TO, $ACT);
}
close(CI);
close(AI);
close($SO);
close($TO);
close($ACT) if defined($ACT);
|