long CSphinxConf::HypConf (search_hyp_t *hyp)
{
search_hyp_t *h;
int32 w1, w2, finishwid, startwid, type[4096];
int32 i, k, t;
finishwid = kb_get_word_id ("</s>");
startwid = kb_get_word_id ("<s>");
if ((finishwid < 0) || (startwid < 0))
return -1;
w1 = finishwid;
w2 = startwid;
type[0] = 3; // Dummy trigram entry before utterance
k = 1;
for (h = hyp; h; h = h->next) {
if (k > 4095)
return -1;
lm_tg_score (w1, w2, h->wid);
type[k++] = lm3g_access_type();
w1 = w2;
w2 = h->wid;
}
lm_tg_score (w1, w2, finishwid);
type[k++] = lm3g_access_type();
type[k++] = 3; // Dummy trigram entries after utterance
for (i = 1, h = hyp; i < k-2; i++, h = h->next) {
t = type[i-1] + type[i] + ((type[i+1] + type[i+2])<<1);
h->conf = (float)((double)(t-6)/12.0);
}
return 0;
}
void vithist_rescore (vithist_t *vh, kbcore_t *kbc,
s3wid_t wid, int32 ef, int32 score, int32 pred, int32 type)
{
vithist_entry_t *pve, tve;
s3lmwid_t lwid;
int32 se, fe;
int32 i;
assert (vh->n_frm == ef);
pve = vh->entry[VITHIST_ID2BLK(pred)] + VITHIST_ID2BLKOFFSET(pred);
/* Create a temporary entry with all the info currently available */
tve.wid = wid;
tve.sf = pve->ef + 1;
tve.ef = ef;
tve.type = type;
tve.valid = 1;
tve.ascr = score - pve->score;
if (pred == 0) { /* Special case for the initial <s> entry */
se = 0;
fe = 1;
} else {
se = vh->frame_start[pve->ef];
fe = vh->frame_start[pve->ef + 1];
}
if (dict_filler_word (kbcore_dict(kbc), wid)) {
tve.lscr = fillpen (kbcore_fillpen(kbc), wid);
tve.score = score + tve.lscr;
tve.pred = pred;
tve.lmstate.lm3g = pve->lmstate.lm3g;
vithist_enter (vh, kbc, &tve);
} else {
lwid = kbcore_dict2lmwid (kbc, wid);
tve.lmstate.lm3g.lwid[0] = lwid;
for (i = se; i < fe; i++) {
pve = vh->entry[VITHIST_ID2BLK(i)] + VITHIST_ID2BLKOFFSET(i);
if (pve->valid) {
tve.lscr = lm_tg_score (kbcore_lm(kbc),
pve->lmstate.lm3g.lwid[1],
pve->lmstate.lm3g.lwid[0],
lwid);
tve.score = pve->score + tve.ascr + tve.lscr;
if ((tve.score - vh->wbeam) >= vh->bestscore[vh->n_frm]) {
tve.pred = i;
tve.lmstate.lm3g.lwid[1] = pve->lmstate.lm3g.lwid[0];
vithist_enter (vh, kbc, &tve);
}
}
}
}
}
/**
* Scores the given N-gram using the given language model.
*
* args:
* wid - the IDs of the sequence of words in the n-gram
* nwd - the number of words in the n-gram
* lm - the language model to use
*
* return: the language model score of the given sequence of words
*/
int
score_ngram(s3lmwid32_t * wid, int nwd, lm_t * lm)
{
int32 score;
score = 0;
if (nwd == 3) {
/* The last argument is a hack: the information there - the dict
* ID - is never used if LM classes are not used, and classes
* are not used in this code. Therefore, the last argument here
* is a nop.
*/
score = lm_tg_score(lm, wid[0], wid[1], wid[2], 0);
}
else if (nwd == 2) {
/* Ditto.
*/
score = lm_bg_score(lm, wid[0], wid[1], 0);
}
else if (nwd == 1) {
/* Ditto.
*/
score = lm_ug_score(lm, wid[0], 0);
}
else {
printf("%d grams not supported\n", nwd);
}
return score;
}
int32 vithist_partialutt_end (vithist_t *vh, kbcore_t *kbc)
{
int32 f, i, b, l;
int32 sv, nsv, scr, bestscore, bestvh;
vithist_entry_t *ve, *bestve;
s3lmwid_t endwid;
lm_t *lm;
/* Find last frame with entries in vithist table */
for (f = vh->n_frm-1; f >= 0; --f) {
sv = vh->frame_start[f]; /* First vithist entry in frame f */
nsv = vh->frame_start[f+1]; /* First vithist entry in next frame (f+1) */
if (sv < nsv)
break;
}
if (f < 0)
return -1;
if (f != vh->n_frm-1){
E_ERROR("No word exits from last frame in block %d\n",vh->n_frm-1);
return -1;
}
/* Terminate in a final </s> node (make this optional?) */
lm = kbcore_lm (kbc);
endwid = lm_finishwid (lm);
bestscore = MAX_NEG_INT32;
bestvh = -1;
for (i = sv; i < nsv; i++) {
b = VITHIST_ID2BLK (i);
l = VITHIST_ID2BLKOFFSET (i);
ve = vh->entry[b] + l;
scr = ve->score;
scr += lm_tg_score (lm, ve->lmstate.lm3g.lwid[1], ve->lmstate.lm3g.lwid[0], endwid);
if (bestscore < scr) {
bestscore = scr;
bestvh = i;
bestve = ve;
}
}
return bestvh;
}
/**
* Scores the given N-gram using the given language model.
*
* args:
* wid - the IDs of the sequence of words in the n-gram
* nwd - the number of words in the n-gram
* lm - the language model to use
*
* return: the language model score of the given sequence of words
*/
int score_ngram(s3lmwid_t *wid, int nwd, lm_t *lm)
{
int32 score, tgscr;
score = 0;
if (nwd == 3) {
score = lm_tg_score(lm, wid[0], wid[1], wid[2]);
} else if (nwd == 2) {
score = lm_bg_score(lm, wid[0], wid[1]);
} else if (nwd == 1) {
score = lm_ug_score(lm, wid[0]);
} else {
printf("%d grams not supported\n", nwd);
}
return score;
}
int32 vithist_utt_end (vithist_t *vh, kbcore_t *kbc)
{
int32 f, i, b, l;
int32 sv, nsv, scr, bestscore, bestvh, vhid;
vithist_entry_t *ve, *bestve=0;
s3lmwid_t endwid;
lm_t *lm;
dict_t *dict;
/* Find last frame with entries in vithist table */
for (f = vh->n_frm-1; f >= 0; --f) {
sv = vh->frame_start[f]; /* First vithist entry in frame f */
nsv = vh->frame_start[f+1]; /* First vithist entry in next frame (f+1) */
if (sv < nsv)
break;
}
if (f < 0)
return -1;
if (f != vh->n_frm-1)
E_ERROR("No word exit in frame %d, using exits from frame %d\n", vh->n_frm-1, f);
/* Terminate in a final </s> node (make this optional?) */
lm = kbcore_lm (kbc);
endwid = lm_finishwid (lm);
bestscore = MAX_NEG_INT32;
bestvh = -1;
for (i = sv; i < nsv; i++) {
b = VITHIST_ID2BLK (i);
l = VITHIST_ID2BLKOFFSET (i);
ve = vh->entry[b] + l;
scr = ve->score;
scr += lm_tg_score (lm, ve->lmstate.lm3g.lwid[1], ve->lmstate.lm3g.lwid[0], endwid);
if (bestscore < scr) {
bestscore = scr;
bestvh = i;
bestve = ve;
}
}
assert (bestvh >= 0);
dict = kbcore_dict (kbc);
if (f != vh->n_frm-1) {
E_ERROR("No word exit in frame %d, using exits from frame %d\n", vh->n_frm-1, f);
/* Add a dummy silwid covering the remainder of the utterance */
assert (vh->frame_start[vh->n_frm-1] == vh->frame_start[vh->n_frm]);
vh->n_frm -= 1;
vithist_rescore (vh, kbc, dict_silwid (dict), vh->n_frm, bestve->score, bestvh, -1);
vh->n_frm += 1;
vh->frame_start[vh->n_frm] = vh->n_entry;
return vithist_utt_end (vh, kbc);
}
/* Create an </s> entry */
vhid = vh->n_entry;
ve = vithist_entry_alloc (vh);
ve->wid = dict_finishwid (dict);
ve->sf = (bestve->ef == BAD_S3FRMID) ? 0 : bestve->ef + 1;
ve->ef = vh->n_frm;
ve->ascr = 0;
ve->lscr = bestscore - bestve->score;
ve->score = bestscore;
ve->pred = bestvh;
ve->type = 0;
ve->valid = 1;
ve->lmstate.lm3g.lwid[0] = endwid;
ve->lmstate.lm3g.lwid[1] = ve->lmstate.lm3g.lwid[0];
return vhid;
}
