Blame view
tools/sctk-2.4.10/src/sclite/ctm2ctm.c
16.2 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 |
#include "sctk.h"
int chop_WTOKE_2(WTOKE_STR1 *ref, WTOKE_STR1 *hyp, int Rstart, int Hstart, int Rendconv, int Hendconv, int max_words, int *Rret_end, int *Hret_end);
SCORES *align_ctm_to_ctm(char *hyp_file, char *ref_file, char *set_title, int feedback, int fcorr, int opt_del, int case_sense, int time_align, int left_to_right, WWL *wwl, char *lm_file){
char *proc = "align_ctm_to_ctm";
WTOKE_STR1 *hyp;
WTOKE_STR1 *ref;
FILE *fp_hyp, *fp_ref;
int hyp_conv_end, hyp_eof;
int ref_conv_end, ref_eof;
int hyp_chop_end, ref_chop_end;
int i;
PATH *path;
int ref_end_chan1, hyp_end_chan1, ref_end_chan2, hyp_end_chan2;
int ref_end, hyp_end, ref_begin, hyp_begin;
int ref_opp_end, hyp_opp_end, ref_opp_begin, hyp_opp_begin;
int dbg = TRUE;
int chunk;
char uttid[100];
int spkr;
int number_of_channels;
SCORES *scor;
NETWORK *hnet, *rnet;
int keep_path = 1;
hyp = WTOKE_STR1_init(hyp_file);
ref = WTOKE_STR1_init(ref_file);
scor = SCORES_init(set_title,20);
scor->ref_fname = (char *)TEXT_strdup((TEXT *)ref_file);
scor->hyp_fname = (char *)TEXT_strdup((TEXT *)hyp_file);
scor->creation_date = get_date();
scor->frag_corr = fcorr;
scor->opt_del = opt_del;
if (wwl != (WWL*)0) {
scor->weight_ali = 1;
scor->weight_file =
(char *)TEXT_strdup((TEXT*)rsprintf("%s Column \"%s\"",wwl->filename,
wwl->weight_desc[0]));
} else if (lm_file != (char *)0) {
scor->weight_ali = 1;
scor->weight_file = (char *)TEXT_strdup((TEXT*)lm_file);
} else
scor->weight_ali = 0;
dbg=0;
if ((fp_hyp = fopen(hyp_file,"r")) == NULL){
fprintf(stderr,"Can't open input hypothesis file %s
",hyp_file);
exit(1);
}
if ((fp_ref = fopen(ref_file,"r")) == NULL){
fprintf(stderr,"Can't open input reference file %s
",ref_file);
exit(1);
}
hyp_eof = ref_eof = FALSE;
do {
if (dbg > 5) printf("Starting do-while loop on conversations"
"--------------------------
");
fill_WTOKE_structure(hyp, fp_hyp, hyp_file, &hyp_eof, case_sense);
fill_WTOKE_structure(ref, fp_ref, ref_file, &ref_eof, case_sense);
if ((hyp->n < hyp->s) && (ref->n > ref->s)) {
fprintf(stderr,"Error: premature EOF in hyp file '%s' at conv"
"ersation '%s' of ref file '%s'
",hyp_file,
ref->word[ref->s].conv,ref_file);
exit(1);
}
if ((hyp->n > hyp->s) && (ref->n < ref->s)) {
fprintf(stderr,"Error: premature EOF in ref file '%s' at conv"
"ersation '%s' of hyp file '%s'
",ref_file,
hyp->word[hyp->s].conv,hyp_file);
exit(1);
}
locate_boundary(ref, ref->s, TRUE, FALSE, &ref_conv_end);
locate_boundary(hyp, hyp->s, TRUE, FALSE, &hyp_conv_end);
/* let's do some checking !!!!!! */
/* Are the conversations in sync */
if (strcmp(ref->word[ref->s].conv,hyp->word[hyp->s].conv)!=0){
fprintf(stderr,"Error: in ref file '%s' and hyp file '%s'
",
ref_file, hyp_file);
fprintf(stderr," file strings out of "
"synchronization around ref-file='%s' hyp-file='%s'
",
ref->word[ref->s].conv,hyp->word[hyp->s].conv);
exit(1);
}
locate_boundary(ref, ref->s, TRUE, TRUE, &ref_end_chan1);
locate_boundary(hyp, hyp->s, TRUE, TRUE, &hyp_end_chan1);
number_of_channels = 1;
hyp_end_chan2 = ref_end_chan2 = -1;
if (ref_end_chan1+1 < ref->n &&
strcmp(ref->word[ref->s].conv,ref->word[ref_end_chan1+1].conv)==0){
locate_boundary(ref, ref_end_chan1+1, TRUE, TRUE, &ref_end_chan2);
number_of_channels = 2;
}
if (hyp_end_chan1+1 < hyp->n &&
strcmp(hyp->word[hyp->s].conv,hyp->word[hyp_end_chan1+1].conv)==0){
locate_boundary(hyp, hyp_end_chan1+1, TRUE, TRUE, &hyp_end_chan2);
number_of_channels = 2;
}
if (number_of_channels == 2){
/* either the ref, hyp or both had two channels. make sure there */
/* are two channels in both */
if (hyp_end_chan2 == -1){
fprintf(stderr,"Error: in ref file '%s' and hyp file '%s'
",
ref_file, hyp_file);
fprintf(stderr," Hyp is missing second channel of file '%s'
",
ref->word[ref->s].conv);
exit(1);
}
if (ref_end_chan2 == -1){
fprintf(stderr,"Error: in ref file '%s' and hyp file '%s'
",
ref_file, hyp_file);
fprintf(stderr," ref is missing second channel of file '%s'
",
hyp->word[hyp->s].conv);
exit(1);
}
} else {
/* There was no second channel */
ref_end_chan2 = ref_end_chan1;
hyp_end_chan2 = hyp_end_chan1;
}
if (feedback >= 1)
printf(" Performing %s alignments for file '%s'.
",
(time_align?"Time-mediated":"Word"),
ref->word[ref->s].conv);
if (dbg > 5){
printf("Located REF boundary: for %s %s -> %d,%d
",
ref->word[ref->s].conv,
ref->word[ref->s].turn,ref->s,ref_conv_end);
printf("Located HYP boundary: for %s %s -> %d,%d
",
hyp->word[hyp->s].conv,
hyp->word[hyp->s].turn,hyp->s,hyp_conv_end);
printf("Located REF Channel 1 boundary: for %d - %d
",
ref->s,ref_end_chan1);
printf("Located HYP Channel 1 boundary: for %d - %d
",
hyp->s,hyp_end_chan1);
printf("Located REF Channel 2 boundary: for %d - %d
",
ref_end_chan1+1,ref_end_chan2);
printf("Located HYP Channel 2 boundary: for %d - %d
",
hyp_end_chan1+1,hyp_end_chan2);
}
/* Start the alignment */
/* Side A */
for (i=0; i<number_of_channels; i++){
int base_len;
chunk = 0;
if (i==0){
ref_end = ref_end_chan1;
hyp_end = hyp_end_chan1;
ref_begin = ref->s;
hyp_begin = hyp->s;
ref_opp_end = ref_end_chan2;
hyp_opp_end = hyp_end_chan2;
ref_opp_begin = ref_end_chan1 + 1;
hyp_opp_begin = hyp_end_chan1 + 1;
} else {
ref_end = ref_end_chan2;
hyp_end = hyp_end_chan2;
ref_begin = ref_end_chan1 + 1;
hyp_begin = hyp_end_chan1 + 1;
ref_opp_end = ref_end_chan1;
hyp_opp_end = hyp_end_chan1;
ref_opp_begin = ref->s;
hyp_opp_begin = hyp->s;
}
if (dbg > 5){
printf("Starting channel %d
",i);
printf("Ref Range: %d - %d
",ref_begin,ref_end);
printf("Hyp Range: %d - %d
",hyp_begin,hyp_end);
}
while (ref_begin<=ref_end || hyp_begin<=hyp_end){
int removals;
base_len = 50;
while (! chop_WTOKE_2(ref, hyp, ref_begin, hyp_begin,
ref_end, hyp_end, base_len,
&ref_chop_end,&hyp_chop_end)){
/* printf("Expanding len from %d to %d
",base_len,
base_len*2); */
base_len *= 2;
}
removals = 0;
/* create the utterance ID */
sprintf(uttid,"(%s-%s-%04d)",ref->word[ref_begin].conv,
ref->word[ref_begin].turn,chunk);
spkr = SCORES_get_grp(scor,
rsprintf("%s-%s",
ref->word[ref_begin].conv,
ref->word[ref_begin].turn));
if (dbg > 5){
printf("Chunk %d: Starting for channel %d id=%s",
chunk,i,uttid);
printf(" Ref Range: %d - %d",ref_begin,ref_chop_end);
printf(" Hyp Range: %d - %d
",hyp_begin,hyp_chop_end);
}
/* create the networks to be aligned */
if ((hnet=Network_create_from_WTOKE(hyp,hyp_begin,
hyp_chop_end,"Hypothesis net",
print_WORD_wt,
equal_WORD2,
release_WORD, null_alt_WORD,
opt_del_WORD,
copy_WORD,make_empty_WORD,
use_count_WORD,
left_to_right))
== NULL_NETWORK){
fprintf(stderr,"%s: Network_create_from_WTOKE failed
",
proc);
return(0);
}
if ((rnet=Network_create_from_WTOKE(ref,ref_begin,
ref_chop_end,"Reference net",
print_WORD_wt,
equal_WORD2,
release_WORD, null_alt_WORD,
opt_del_WORD,
copy_WORD,make_empty_WORD,
use_count_WORD,
left_to_right))
== NULL_NETWORK){
fprintf(stderr,"%s: Network_create_from_WTOKE failed
",
proc);
return(0);
}
path = network_dp_align_texts((TEXT *)0, rnet,
(TEXT *)0, hnet,
FALSE, case_sense,
(char *)uttid, fcorr,
opt_del, time_align, wwl, lm_file);
BF_SET(path->attrib,PA_HYP_WTIMES);
BF_SET(path->attrib,PA_REF_WTIMES);
if (ref->has_conf) BF_SET(path->attrib,PA_REF_CONF);
if (hyp->has_conf) BF_SET(path->attrib,PA_HYP_CONF);
sort_PATH_time_marks(path);
if (dbg > 10){
printf("After Sorting
");
PATH_print_wt(path, stdout);
}
add_PATH_score(scor,path,spkr, keep_path);
PATH_add_file(path,ref->word[ref_begin].conv);
PATH_add_channel(path,ref->word[ref_begin].turn);
/* PATH_add_label(path,(char *)stm->seg[rs].labels); */
ref_begin = ref_chop_end+1;
hyp_begin = hyp_chop_end+1;
chunk++;
}
}
hyp->s = hyp_conv_end+1;
ref->s = ref_conv_end+1;
} while (hyp->s <= hyp->n || ref->s <= ref->n);
free_mark_file(ref);
free_mark_file(hyp);
fclose(fp_hyp);
fclose(fp_ref);
return (scor);
}
/*******************************************************************/
/* Compute the overlap of two segments in time and return that */
/* time. Negative times indicate no overlap. There are 6 cases */
/* Case 1: S1 t1 t2 */
/* S2 t1 t2 */
/* */
/* Case 2: S1 t1 t2 */
/* S2 t1 t2 */
/* */
/* Case 3a: S1 t1 t2 */
/* S2 t1 t2 */
/* */
/* Case 3b: S1 t1 t2 */
/* S2 t1 t2 */
/* */
/* Case 4a: S1 t1 t2 */
/* S2 t1 t2 */
/* */
/* Case 4b: S1 t1 t2 */
/* S2 t1 t2 */
/* */
double overlap(double s1_t1, double s1_t2, double s2_t1, double s2_t2)
{
double rval;
char *rule;
int dbg=0;
/* Case 1: */
if (s1_t2 < s2_t1)
rule="Case 1", rval=(s1_t2 - s2_t1);
/* Case 2: */
if (s1_t1 > s2_t2)
rule="Case 2", rval=(s2_t2 - s1_t1);
/* Case 3: */
if (s1_t1 < s2_t1){
/* Case 3a: */
if (s1_t2 < s2_t2)
rule="Case 3a", rval=(s1_t2 - s2_t1);
else /* Case 3b: */
rule="Case 3b", rval=(s2_t2 - s2_t1);
} else { /* Case 4: */
/* Case 4a: */
if (s1_t2 < s2_t2)
rule="Case 4a", rval=(s1_t2 - s1_t1);
else /* Case 4b: */
rule="Case 4b", rval=(s2_t2 - s1_t1);
}
if (dbg){
printf("Overlap: s1_t1=%2.2f s1_t2=%2.2f",s1_t1,s1_t2);
printf(" s2_t1=%2.2f s2_t2=%2.2f Rule:%s Rval=%2.2f
",
s2_t1,s2_t2,rule,rval);
}
return(rval);
}
int chop_WTOKE_2(WTOKE_STR1 *ref, WTOKE_STR1 *hyp, int Rstart, int Hstart, int Rendconv, int Hendconv, int max_words, int *Rret_end, int *Hret_end)
{
int Rend, Hend, Rend_sync, Hend_sync;
char *proc="chop_WTOKE_2";
int dbg=0, skipped_alt;
double Rgap_t1, Rgap_t2, Hgap_t1, Hgap_t2;
if (dbg)
printf("%s: Start Ref=%d Start Hyp=%d Max_words=%d
",
proc,Rstart,Hstart,max_words);
/* Find the end of this segment */
/* locate_boundary(ref, Rstart, TRUE, TRUE, &Rendconv);
locate_boundary(hyp, Hstart, TRUE, TRUE, &Hendconv); */
if (dbg) printf("%s: Conversation Rendconv=%d Hendconv=%d
",
proc,Rendconv,Hendconv);
if ((Rstart > Rendconv) && (Hstart > Hendconv))
return(0);
/* if we've hit the end of the conversation, send everything */
if (((Rendconv - Rstart) <max_words) && ((Hendconv - Hstart) < max_words)){
*Rret_end = Rendconv;
*Hret_end = Hendconv;
if (dbg) printf("%s: R1 Success Chop Ref (%d-%d) Hyp (%d-%d)
",
proc,Rstart,*Rret_end,Hstart,*Hret_end);
return(1);
}
/* set the indexes to maximum length */
Rend = MIN(Rstart + max_words, Rendconv);
Hend = MIN(Hstart + max_words, Hendconv);
if (dbg) printf("%s: Limited Rend=%d Hend=%d
",proc,Rend,Hend);
/* synchronize the ends to permit searching */
while (Rend >= Rstart && ref->word[Rend].alternate)
Rend--;
skipped_alt = TRUE;
while (skipped_alt){
skipped_alt = FALSE;
if (ref->word[Rend].t1 > hyp->word[Hend].t1)
while (!skipped_alt &&
Rend >= Rstart &&
ref->word[Rend].t1 > hyp->word[Hend].t1 &&
overlap(ref->word[Rend].t1,
ref->word[Rend].t1+ref->word[Rend].dur,
hyp->word[Hend].t1,
hyp->word[Hend].t1+hyp->word[Hend].dur) < 0.0)
for (Rend--; Rend >= Rstart && ref->word[Rend].alternate; ) {
Rend--;
skipped_alt = TRUE;
}
else if (ref->word[Rend].t1 < hyp->word[Hend].t1)
while (Hend >= Hstart &&
ref->word[Rend].t1 < hyp->word[Hend].t1 &&
overlap(ref->word[Rend].t1,
ref->word[Rend].t1+ref->word[Rend].dur,
hyp->word[Hend].t1,
hyp->word[Hend].t1+hyp->word[Hend].dur) < 0.0)
Hend --;
}
if (dbg) {
printf("%s: Synchronized Rend=%d Hend=%d ",proc,Rend,Hend);
printf("Times: %2.2f,%2.2f %2.2f,%2.2f
",
ref->word[Rend].t1, ref->word[Rend].t1+ref->word[Rend].dur,
hyp->word[Hend].t1, hyp->word[Hend].t1+hyp->word[Hend].dur);
}
Rend_sync = Rend;
Hend_sync = Hend;
/* begin the backward search */
while (Rend > Rstart && Hend > Hstart){
/* first compute t1 and t2 for the gap in each segment */
if (Rend >= Rstart){
Rgap_t1 = ref->word[Rend].t1 + ref->word[Rend].dur;
Rgap_t2 = (Rend == Rendconv) ? 999999.99 : ref->word[Rend+1].t1;
} else {
Rgap_t1 = 0.0;
Rgap_t2 = ref->word[Rstart].t1;
}
if (Hend >= Hstart) {
Hgap_t1 = hyp->word[Hend].t1 + hyp->word[Hend].dur;
Hgap_t2 = (Hend == Hendconv) ? 999999.99 : hyp->word[Hend+1].t1;
} else {
Hgap_t1 = 0.0;
Hgap_t2 = hyp->word[Hstart].t1;
}
if (dbg) {
printf("
%s: Gaps: ref %2.2f,%2.2f hyp %2.2f,%2.2f
",
proc,Rgap_t1, Rgap_t2, Hgap_t1, Hgap_t2);
printf("%s: **** middle word is Rend or Hend
",proc);
dump_word_tokens2(ref,Rend-1,Rend+1);
dump_word_tokens2(hyp,Hend-1,Hend+1);
}
/* check to see if there is any overlap */
/* yes if : S1 | | # | | */
/* S2 | | # | | */
if (overlap(Rgap_t1,Rgap_t2,Hgap_t1,Hgap_t2) < 0.0){
/* no overlap decide which to move */
if (dbg) printf("%s: Shifting down
",proc);
skipped_alt = TRUE;
while (skipped_alt){
skipped_alt = FALSE;
if (ref->word[Rend].t1 > hyp->word[Hend].t1){
if (Rend > Rstart) {
Rend --;
while (Rend >= Rstart && ref->word[Rend].alternate) {
Rend--;
skipped_alt = TRUE;
}
}
if (dbg) printf("%s: Moving Ref
",proc);
while (!skipped_alt &&
Hend >= Hstart &&
ref->word[Rend].t1 < hyp->word[Hend].t1 &&
overlap(ref->word[Rend].t1,
ref->word[Rend].t1+ref->word[Rend].dur,
hyp->word[Hend].t1,
hyp->word[Hend].t1+hyp->word[Hend].dur)< 0.0){
Hend --;
if (dbg) printf("%s: Adjusting Hyp
",proc);
}
} else if (ref->word[Rend].t1 <= hyp->word[Hend].t1) {
if (Hend > Hstart)
Hend --;
if (dbg) printf("%s: Moving Hyp
",proc);
while (!skipped_alt &&
Rend >= Rstart &&
ref->word[Rend].t1 > hyp->word[Hend].t1 &&
overlap(ref->word[Rend].t1,
ref->word[Rend].t1+ref->word[Rend].dur,
hyp->word[Hend].t1,
hyp->word[Hend].t1+ref->word[Hend].dur)<0.0){
for (Rend--;
Rend >= Rstart && ref->word[Rend].alternate; ) {
Rend--;
skipped_alt = TRUE;
}
if (dbg) printf("%s: Adjusting Ref
",proc);
}
}
}
} else { /* Overlap exists !!!!! Return the values */
*Rret_end = Rend;
*Hret_end = Hend;
if (dbg) printf("%s: R2 Success Chop Ref (%d-%d) Hyp (%d-%d)
",
proc,Rstart,*Rret_end,Hstart,*Hret_end);
return(1);
}
}
if (Rend > Rstart || Hend > Hstart){
*Rret_end = Rend;
*Hret_end = Hend;
if (dbg) printf("%s: R3 Success Chop Ref (%d-%d) Hyp (%d-%d)
",
proc,Rstart,*Rret_end,Hstart,*Hret_end);
return(1);
}
/* fprintf(stdout,"%s: Failure to find chop'd segment, returning 0
",proc); */
return(0);
}
|