Yannick Estève / ONTRAC-Kaldi

sclite.htm 17.1 KB
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 
<!-- $Id: sclite.htm,v 1.2 2004/07/27 15:07:09 jfiscus Exp $ -->
<HTML>
<HEAD>
<CENTER><TITLE>NIST SCLITE Scoring Package Version 1.5</TITLE>
</HEAD>
<BODY></CENTER><p><hr>

<H1> 
<A NAME="sclite_name_0">
NAME</A>
</H1>
sclite - score speech recognition system output
<p><p><hr>

<H1> 
<A NAME="sclite_synopsis_0">
SYNOPSIS</A>
</H1>
<STRONG>sclite 
<a href="options.htm#option_r_name_0">-r</a> reffile [ fmt ] 
<a href="options.htm#option_h_name_0">-h</a> hypfile [ fmt [ title ] ] 
<a href="options.htm#options_name_0">OPTIONS</a>
</STRONG>
<p>
<H1> <A NAME="sclite_description_0">DESCRIPTION</A></H1>
<p>
The program <A HREF="sclite.htm#sclite_name_0">sclite</A> is a tool
for scoring and evaluating the output of speech recognition systems.
Sclite is part of the <A HREF="sctk.htm">NIST SCTK</A> Scoring Tookit.
The program compares the hypothesized text (HYP) output by the speech
recognizer to the correct, or reference (REF) text.  After comparing
REF to HYP, (a process called 
<A HREF="sclite.htm#sclite_alignment_process_0">alignment</A>),
statistics are gathered during the <A HREF="sclite.htm#scoring_process">scoring process</A>
 and a variety of 
<A HREF="outputs.htm#outputs_name_0">reports</A> can be produced to
summarize the performance of the recognition system.

<H1> 
<A NAME="sclite_alignment_process_0">THE ALIGNMENT PROCESS</A>
</H1>
<p>

The Alignment process consists of two steps: 1) selecting matching
REF and HYP texts, and 2) performing an alignment of the reference 
and hypothesis texts.   
<p>
<strong> Step 1: Selection of matching REF and HYP texts </strong>
<UL>
<A HREF="sclite.htm#sclite_name_0">Sclite</A>
accepts as input a wide variety of file formats.  The type of input
formats define the algorithm for selecting matching REF and 
HYP texts.  Currently sclite uses four algorithms:
<UL>
<p>
<strong> Utterance ID Matching: </strong>
<UL>

Input reference and hypothesis files in 
"<a href="infmts.htm#trn_fmt_name_0">trn</a>" transcript format can be
aligned by either dynamic programming 
(<a href="sclite.htm#dynam_prog_0">DP</a>) or 
<a href="sclite.htm#gnu_diff_alignment_0">GNU's "diff"</a>.
<p> 
When alignments are performed via DP, corresponding REF and HYP
records with the same utterance id's are located in the REF and HYP
files.  DP Alignment and scoring are then performed on each pair of
records.  Only the utterance ID's present in the HYP file are aligned and
scored.  This means the REF file may contain more utterance 
records than the HYP.
<p>  
When "diff" is used for alignment, corresponding REF and HYP records
with the same utterance id's are located in the REF and HYP files.
Rather than execute "diff" for each pair of records, all matching REF
and HYP pairs are re-formatted to be newline separated words and
written to a temporary files.  Using the two temporary files, "diff"
is then called to perform a global alignment.  The output of "diff" is
re-chunking into REF/HYP records by applying the rule: include all
words in the output stream up to and including the last word in the
reference record.
<p>  
The reference file can contain extra transcripts, only needed
transcripts are loaded.

</UL>

<p>

<STRONG> Word Time Mark Matching: </STRONG>

<UL>
When both the REF and HYP files are in the 
"<a href="infmts.htm#ctm_fmt_name_0">ctm</a>" format,
The first step in the alignment process is to segment both the
reference and hypothesis word lists by locating common areas of
silence, (i.e.  the absence of a word time mark).  Once completed, the
resulting "segments" are aligned via dynamic programming and scored as
usual.
<p>
By default, the DP alignment is performed using word-to-word distances
measures of: 0, 3, 3, 4 for correct, insertions, deletions and
substitutions respectively.
<p> 
Optionally, the command line flag 
'<a href="options.htm#option_T_name_0">-T</a>' forces the alignments to be
performed using <A HREF="sclite.htm#time-mediated">time-mediated</A> alignments.

</UL>

<p>

<STRONG> Reference Segment Time Mark to Hypothesis Word Time Mark </STRONG>

<UL>
When the reference file format is "<a href="infmts.htm#stm_fmt_name_0">stm</a>"
and the hypothesis file format is
"<a href="infmts.htm#ctm_fmt_name_0">ctm</a>", sclite chops up the hypothesis file into regions matching
the reference segments.  Currently, there a two methods of chopping
the hypothesis file.  The method is dependent on the text alignment algorithm. 

<p>
 
When DP alignments are performed, the hypothesis file is segmented to
match the reference segments by selecting the string of hypothesized
words whose times occur before the end of each reference segment.  The
midpoint time of a word is used to determine if the word falls within
a segment.  DP alignments are then performed on the selected
hypothesis words and the reference segment.

<p> 
 
If the alignments are performed via "diff", pre-process the input
reference and hypothesis texts, creating temporary reference and
hypothesis files with one word per line.  Then use GNU's "diff"
program to perform a global alignment on the word lists.  The output
of "diff" is re-chunked into segments for scoring.  Alternate
reference transcripts can not be used with "diff" alignments.
 
</UL>
<p>
<STRONG> Reference Segment Time Mark to Hypothesis Text file </STRONG>
<ul>

When the reference file format 
"<a href="infmts.htm#stm_fmt_name_0">stm</a>" and the hypothesis file
format "<a href="infmts.htm#txt_fmt_name_0">txt</a>" are used as
inputs, the same alignment and scoring algorithm is used as describe
above under the label "Reference Segment Time Mark to Hypothesis Word
Time Mark" by GNU diff alignments.

</ul>
</UL>


</UL>
<p>

<strong> Step 2: Text Alignments </strong>

<UL>
<A HREF="sclite.htm#sclite_name_0">Sclite</A>
can use either of two algorithms for finding alignments
between reference and hypothesis word strings.  The first, and most
widely accepted, uses dynamic programming (DP) and the second uses
GNU's "diff", a FSF (Free Software Foundation) program for comparing
text files.
<p>
<A name="dynam_prog_0">Dynamic Programming string alignment:</a>
<UL>
<p>
The DP string alignment algorithm performs a global minimization of a
Levenshtein distance function which weights the cost of correct words,
insertions, deletions and substitutions as 0, 3, 3 and 4 respectively.
The computational complexity of DP is 0(NN).
<p>
When evaluating the output of speech recognition systems, the
precision of generated statistics is directly correlated to the
reference text accuracy.  But uttered words can be coarticulated or
mumbled to where they have ambiguous transcriptions, (e.i., "what are"
or "what're").  In order to more accurately represent ambiguous
transcriptions, and not penalize recognition systems, the ARPA
community agreed upon a format for specifying alternative reference
transcriptions.  The convention, when used on the case above, allows
the recognition system to output either transcripts, "what are" or
"what're", and still be correct.
<p>
The case above handles ambiguously spoken words which are loud enough
for the transcriber to think something should be recognized.  For
mumbled or quietly spoken words, the ARPA community agreed to neither
penalize systems which correctly recognized the word, nor penalize
systems which did not.  To accommodate this, a NULL word, "@", can be
added to an alternative reference transcript.  For example, "the" is
often spoken quickly with little acoustic evidence.  If "the" and "@"
are alternates, the recognition system will be given credit for
outputting "the" but not penalized if it does not.
<p>
The presence of alternate transcriptions represents added
computational complexity to the DP algorithm.  Rather than align all
alternate reference texts to the hypothesis text, then choose the
lowest error rate alignment, this implementation of DP aligns two word
networks, thus reducing the computational complexity from 2^(ref_alts +
hyp_alts) * O(N_ref * N_hyp) to O((N_ref+ref_alts) *
(N_hyp+hyp_alts)).
<p>
<UL>
For a detailed explanation of DP alignment, see TIME WARPS, STRING
EDITS, AND MACROMOLECULES: THE THEORY AND PRACTICE OF SEQUENCE
COMPARISON, by Sankoff and Kruskal, ISBN 0-201-07809-0.
</UL>
<p>

As noted above, DP alignment minimizes a distance function that is applied
to word pairs.  In addition to the "word" alignments which uses
a distance function defined by static weights, the sclite DP alignment module can
use two other distance functions.  The first, called <B> Time-Mediated</B> alignment
and the second called <B> Word-Weight-Mediated</B> alignment.
<P>

<A NAME="time-mediated"><B>Time-Mediated Alignment</B></A>
<ul>

Time-Mediated alignment is a variation of DP alignment where
word-to-word distances are based on the time of occurence for
individual words.  Time-mediated alignments are performed when the '<a
href="options.htm#option_T_name_0">-T</a>' option is exercised and the
input formats for both the reference and hypothesis files are in "<a
href="infmts.htm#ctm_fmt_name_0">ctm</a>" format.

<p>

Time-mediated alignments are computed by replacing the
standard word-to-word distance weights of 0, 3, 3, and 4 with measures
based on beginning and ending word times.  The formulas for
time-mediated word-to-word distances are:
<p>

<ul>
D(correct) = | T1(ref) - T1(hyp) | + | T2(ref) - T2(hyp) |
<br>
D(insertion)  = T2(hyp) - T1(hyp)
<br>
D(deletion)  = T2(ref) - T1(ref)
<br>
D(substitution) = | T1(ref) - T1(hyp) | + | T2(ref) - T2(hyp) | + 0.001
<br>
 Distance for an Insertion or Deletion of the NULL Token '@' = 0.001
<p>
               Where,
	<ul>
                   T1(x) is the beginning time mark of word x <br>
                    T2(x) is the ending time mark of word x
	</ul>
</ul>
<P>
</UL>

<A NAME="word-weight-mediated"><B>Word-Weight-Mediated Alignment</B></A>
<UL>

Word-weight-mediated alignment is a variation of DP alignments where word-to-word distances 
are based on pre-defined word-weights.  Each word has a unique weight assigned
to it, via either a word-weight-list file, using the <a
href="options.htm#option_w_name_0">-w</a> option, or through a
language model file, using the <a href="options.htm#option_L_name_0">
-L</a> option.

The formulas for word-weight-mediated word-to-word distances are:
<p>

<ul>
D(correct) = 0.0
<br>
D(insertion)  = W(hyp)
<br>
D(deletion)  = W(ref)
<br>
D(substitution) = W(hyp) + W(ref)
<br>
 Distance for and Insertion or Deletion of the NULL Token '@' = 0.001
<p>
<UL>
               Where W(x) is the weight assigned to word 'x'.
</ul>

</UL>
</UL>
</UL>
<p>
<A name="gnu_diff_alignment_0"></a>
String alignments via GNU's "diff":
<UL>
<p>
While the DP algorithm has the advantage of flexibility, it is slow
for aligning large chunks of text.  To address the speed concerns, an
alternative string alignment module, which utilizes GNU's "diff", has
been added to sclite.  The sclite program pre-processes the input
reference and hypothesis texts, creating temporary reference and
hypothesis files with one word per line.  Then GNU's "diff" program is
used to perform a global alignment on the word lists and the output is
re-chunked into utterances or text segments for scoring.
<p>

Alignments can be performed with "diff" in about half the time taken
for DP alignments on the standard 300 Utterance ARPA CSRNAB test set.
However, in the opinion of the author, "diff" has the following bad
effects:

<UL>
<p>
1. it can not accommodate transcription alternations,
<p>
2. "diff" does not produce the same alignments as the DP alignments,
<p>
3. there is an increase measured error rates.
</UL>
</UL>
</UL>

<A name="scoring_process"> <H1> THE SCORING PROCESS </H1> </A>

<UL>
After reference and hypothesis texts have been aligned, scores are
tallied for each speaker and each ref/hyp pair.  After the tallies
are made, a variety if output reports are generated by using the 
'<A HREF="options.htm#option_o_name_0">-o</a>' option.  
Here is a set of <a href="outputs.htm#output_reports_name_0">examples</a>.

<P>
The categories tallied are: 
<CENTER>
<Table>

<TR> <TD> Percent of correct words 
     <TD> = 
     <TD> <U> # Correct words </U> <br> # Reference words  <TD> * 100
<TR> <TD> Percent of substituted words
     <TD> = 
     <TD> <U> # Substituted words </U> <BR> # Reference words <TD> * 100
<TR> <TD> Percent of inserted words 
     <TD> = 
     <TD> <U> # Inserted words </U> <BR> # Reference words <TD> * 100
<TR> <TD> Percent of deleted words 
     <TD> = 
     <TD> <U> # Deleted words </U> <BR> # Reference words <TD> * 100
<TR>
<TR> <TD> Percent of sentence errors 
     <TD> = 
     <TD> <U> # incorrect ref and hyp pairs </U> <BR> # ref and hyp pairs <TD> * 100
</TABLE>
</CENTER>

<P> A variation in scoring called <A NAME="weighted-word-scoring"> <B>
Weighted-Word Scoring </B> </A> can also be implemented by sclite.
After <A HREF="sclite.htm#word-weight-mediated"> Word-Weight-Mediated
Alignment</A>, the word weights can be tallied to produce
weighted-word scores.  The formulas for weighted-word scoring are very
simliar to word scoring described above.  The difference is rather than
assume each word has the same weight, 1 in the case of word scoring, 
each individual word has a different weight.  The word scoring formulas become:

<CENTER>
<Table>

<TR> <TD> Weighted Percent of correct words
     <TD> = 
     <TD> <U> Sum of W(hyp) if correct </U> <br> Sum of W(ref)  <TD> * 100
<TR> <TD> Weighted Percent of substituted words
     <TD> = 
     <TD> <U> Sum of W(hyp) + W(ref) if substituted </U> <BR> Sum of W(ref)  <TD> * 100
<TR> <TD> Weighted Percent of inserted words 
     <TD> = 
     <TD> <U> Sum of W(hyp) if inserted </U> <BR> Sum of W(ref)  <TD> * 100
<TR> <TD> Weighted Percent of deleted words 
     <TD> = 
     <TD> <U> Sum of W(ref) if deleted </U> <BR> Sum of W(ref) <TD> * 100
</TABLE>
</CENTER>
<UL>
W(hyp) is the weight assigned to a hypothesis word, and W(ref) is the weight
assigned to a reference word.  Optionally deletable words have the default
weight of 0.0.
</UL>

</ul>
<H1> 
<A NAME="graphs_name_0">
WORD CONFIDENCE MEASURE EVALUATION</A>
</H1>
<ul>

Confidence scores for each hypothesized word were requested of the
LVCSR (Large Vocabulary Speech Recognition) participants beginning
with the April 1996 evaluation.  Each site was asked to do its analysis
of these scores which were not processed by NIST.  A review meeting
was held at NIST in August 1996 which resulted in a decision to apply
an agreed upon standard metric.

<p>

Confidence scores as they have been implemented are associated with
each hypothesized word.  (The issue has been raised whether for
languages such as Mandarin, where character error rate is considered
the primary measure of performance, the confidence ought to be
associated with characters.)  The confidence score p<sub>c</sub>,
associated with a word must be in the closed interval [0,1] and
presumably, given the entropy related metric defined below, in the
open interval (0,1).  It should represent the system's best estimate
of the a posterior probability that the hypothesized word is
correct. (Correct here necessarily is with respect to an alignment
procedure of the reference and hypothesis word strings.)

<p>

A single metric to use in the evaluartion of confidence scores was
adopted at the August meeting.  This is a normalized version of the
cross entropy or mutual information.  Specifically, the metric is
defined as:

<p>
<center><IMG SRC="equation.jpg"></Center>
<p>

Sclite will automatically detect the presence of confidence measures 
when reading in a hypothesis "<a href="infmts.htm#ctm_fmt_name_0">ctm</a>"
file.  When sclite detects the confidence scores, the report genererated
by the options "<a href="options.htm#option_o_name_0">-o sum</a>" has 
an additional column containing the Normalized Cross Entropy (NCE).

<p>

Output graphs concerning confidence estimates are generated by using the 
'<A HREF="options.htm#option_C_name_0">-C</a>'
option.  A variety of graphs can be created:

<ol>
<li> DET Curve <a href="outputs.htm#outputs_det_name_0"> Example </a>
<li> Binned Histogram<a href="outputs.htm#outputs_bhist_name_0"> Example </a>
<li> Word Confidence Score Histogram
	<a href="outputs.htm#outputs_hist_name_0"> Example </a>
</ol>

	
</ul> 
<h1> REVISION HISTORY </h1>
<ul>
See <a href="revis.htm#revisions_name_0">revision.txt</a> in
the main directory of the sclite source code directory
package.  
</ul>
<h1> EXAMPLE USES OF <a href="sclite.htm#sclite_name_0">SCLITE</a> </h1>
<ul>
The <a href="sclite.htm#sclite_name_0">sclite</a> scoring utility was
written to be used as a standard scoring tool for the ARPA speech
recognition benchmark tests.  Since evaluation paradigms have changed 
over the past several years, file formats and scoring proceedures have
changed as well.  This utility supports the following speech recognition
benchmark tests:
<ul>
Utterance based evaluations:
	<ul>
	Resource Management
	<br>
	ATIS (Airline Travel Information Systems): 
	</ul>
Found speech evaluations:
	<ul>
	Hub 4 - Marketplace and Broadcast News
	<br>
	Hub 5 - LVCSR Switchboard 
	</ul>
</ul>
</ul>
<h1> BUGS/COMMENTS </h1>
<ul>

Please contact Jon Fiscus at NIST with any bug reports or comments at
the email address 
<A HREF="mailto:jonathan.fiscus@nist.gov">jonathan.fiscus@nist.gov </A> or
by phone, (301)-975-3182.  Please include the version number of rover,
and any other relevant information such as OS, compiler, etc.  </ul>
</ul>

</BODY>
</HTML>