ps_nbest_t *
ps_nbest(ps_decoder_t *ps, int sf, int ef,
char const *ctx1, char const *ctx2)
{
ps_lattice_t *dag;
ngram_model_t *lmset;
ps_astar_t *nbest;
float32 lwf;
int32 w1, w2;
if (ps->search == NULL)
return NULL;
if ((dag = ps_get_lattice(ps)) == NULL)
return NULL;
/* FIXME: This is all quite specific to N-Gram search. Either we
* should make N-best a method for each search module or it needs
* to be abstracted to work for N-Gram and FSG. */
if (0 != strcmp(ps_search_name(ps->search), PS_SEARCH_NGRAM)) {
lmset = NULL;
lwf = 1.0f;
} else {
lmset = ((ngram_search_t *)ps->search)->lmset;
lwf = ((ngram_search_t *)ps->search)->bestpath_fwdtree_lw_ratio;
}
w1 = ctx1 ? dict_wordid(ps_search_dict(ps->search), ctx1) : -1;
w2 = ctx2 ? dict_wordid(ps_search_dict(ps->search), ctx2) : -1;
nbest = ps_astar_start(dag, lmset, lwf, sf, ef, w1, w2);
return (ps_nbest_t *)nbest;
}
const char*
ps_get_kws(ps_decoder_t *ps, const char* name)
{
ps_search_t *search = ps_find_search(ps, name);
if (search && strcmp(PS_SEARCH_KWS, ps_search_name(search)))
return NULL;
return search ? kws_search_get_keywords(search) : NULL;
}
fsg_model_t *
ps_get_fsg(ps_decoder_t *ps, const char *name)
{
ps_search_t *search = ps_find_search(ps, name);
if (search && strcmp(PS_SEARCH_FSG, ps_search_name(search)))
return NULL;
return search ? ((fsg_search_t *) search)->fsg : NULL;
}
ngram_model_t *
ps_get_lm(ps_decoder_t *ps, const char *name)
{
ps_search_t *search = ps_find_search(ps, name);
if (search && strcmp(PS_SEARCH_NGRAM, ps_search_name(search)))
return NULL;
return search ? ((ngram_search_t *) search)->lmset : NULL;
}
fsg_set_t *
ps_get_fsgset(ps_decoder_t *ps)
{
if (ps->search == NULL
|| 0 != strcmp(ps_search_name(ps->search), "fsg"))
return NULL;
return (fsg_set_t *)ps->search;
}
ngram_model_t *
ps_get_lmset(ps_decoder_t *ps)
{
if (ps->search == NULL
|| 0 != strcmp(ps_search_name(ps->search), "ngram"))
return NULL;
return ((ngram_search_t *)ps->search)->lmset;
}
static ps_search_t *
ps_find_search(ps_decoder_t *ps, char const *name)
{
gnode_t *gn;
for (gn = ps->searches; gn; gn = gnode_next(gn)) {
if (0 == strcmp(ps_search_name(gnode_ptr(gn)), name))
return (ps_search_t *)gnode_ptr(gn);
}
return NULL;
}
static int
set_search_internal(ps_decoder_t *ps, ps_search_t *search)
{
ps_search_t *old_search;
if (!search)
return -1;
search->pls = ps->phone_loop;
old_search = (ps_search_t *) hash_table_replace(ps->searches, ps_search_name(search), search);
if (old_search != search)
ps_search_free(old_search);
return 0;
}
int
ps_set_search(ps_decoder_t *ps, const char *name)
{
ps_search_t *search = ps_find_search(ps, name);
if (search)
ps->search = search;
/* Set pl window depending on the search */
if (!strcmp(PS_SEARCH_NGRAM, ps_search_name(search))) {
ps->pl_window = cmd_ln_int32_r(ps->config, "-pl_window");
} else {
ps->pl_window = 0;
}
return search ? 0 : -1;
}
int
ps_add_word(ps_decoder_t *ps,
char const *word,
char const *phones,
int update)
{
int32 wid;
s3cipid_t *pron;
hash_iter_t *search_it;
char **phonestr, *tmp;
int np, i, rv;
/* Parse phones into an array of phone IDs. */
tmp = ckd_salloc(phones);
np = str2words(tmp, NULL, 0);
phonestr = ckd_calloc(np, sizeof(*phonestr));
str2words(tmp, phonestr, np);
pron = ckd_calloc(np, sizeof(*pron));
for (i = 0; i < np; ++i) {
pron[i] = bin_mdef_ciphone_id(ps->acmod->mdef, phonestr[i]);
if (pron[i] == -1) {
E_ERROR("Unknown phone %s in phone string %s\n",
phonestr[i], tmp);
ckd_free(phonestr);
ckd_free(tmp);
ckd_free(pron);
return -1;
}
}
/* No longer needed. */
ckd_free(phonestr);
ckd_free(tmp);
/* Add it to the dictionary. */
if ((wid = dict_add_word(ps->dict, word, pron, np)) == -1) {
ckd_free(pron);
return -1;
}
/* No longer needed. */
ckd_free(pron);
/* Now we also have to add it to dict2pid. */
dict2pid_add_word(ps->d2p, wid);
/* TODO: we definitely need to refactor this */
for (search_it = hash_table_iter(ps->searches); search_it;
search_it = hash_table_iter_next(search_it)) {
ps_search_t *search = hash_entry_val(search_it->ent);
if (!strcmp(PS_SEARCH_NGRAM, ps_search_name(search))) {
ngram_model_t *lmset = ((ngram_search_t *) search)->lmset;
if (ngram_model_add_word(lmset, word, 1.0) == NGRAM_INVALID_WID) {
hash_table_iter_free(search_it);
return -1;
}
}
if (update) {
if ((rv = ps_search_reinit(search, ps->dict, ps->d2p) < 0)) {
hash_table_iter_free(search_it);
return rv;
}
}
}
/* Rebuild the widmap and search tree if requested. */
return wid;
}
int
ps_reinit(ps_decoder_t *ps, cmd_ln_t *config)
{
const char *path;
const char *keyphrase;
int32 lw;
if (config && config != ps->config) {
cmd_ln_free_r(ps->config);
ps->config = cmd_ln_retain(config);
}
err_set_debug_level(cmd_ln_int32_r(ps->config, "-debug"));
ps->mfclogdir = cmd_ln_str_r(ps->config, "-mfclogdir");
ps->rawlogdir = cmd_ln_str_r(ps->config, "-rawlogdir");
ps->senlogdir = cmd_ln_str_r(ps->config, "-senlogdir");
/* Fill in some default arguments. */
ps_init_defaults(ps);
/* Free old searches (do this before other reinit) */
ps_free_searches(ps);
ps->searches = hash_table_new(3, HASH_CASE_YES);
/* Free old acmod. */
acmod_free(ps->acmod);
ps->acmod = NULL;
/* Free old dictionary (must be done after the two things above) */
dict_free(ps->dict);
ps->dict = NULL;
/* Free d2p */
dict2pid_free(ps->d2p);
ps->d2p = NULL;
/* Logmath computation (used in acmod and search) */
if (ps->lmath == NULL
|| (logmath_get_base(ps->lmath) !=
(float64)cmd_ln_float32_r(ps->config, "-logbase"))) {
if (ps->lmath)
logmath_free(ps->lmath);
ps->lmath = logmath_init
((float64)cmd_ln_float32_r(ps->config, "-logbase"), 0,
cmd_ln_boolean_r(ps->config, "-bestpath"));
}
/* Acoustic model (this is basically everything that
* uttproc.c, senscr.c, and others used to do) */
if ((ps->acmod = acmod_init(ps->config, ps->lmath, NULL, NULL)) == NULL)
return -1;
if (cmd_ln_int32_r(ps->config, "-pl_window") > 0) {
/* Initialize an auxiliary phone loop search, which will run in
* "parallel" with FSG or N-Gram search. */
if ((ps->phone_loop =
phone_loop_search_init(ps->config, ps->acmod, ps->dict)) == NULL)
return -1;
hash_table_enter(ps->searches,
ckd_salloc(ps_search_name(ps->phone_loop)),
ps->phone_loop);
}
/* Dictionary and triphone mappings (depends on acmod). */
/* FIXME: pass config, change arguments, implement LTS, etc. */
if ((ps->dict = dict_init(ps->config, ps->acmod->mdef, ps->acmod->lmath)) == NULL)
return -1;
if ((ps->d2p = dict2pid_build(ps->acmod->mdef, ps->dict)) == NULL)
return -1;
lw = cmd_ln_float32_r(config, "-lw");
/* Determine whether we are starting out in FSG or N-Gram search mode.
* If neither is used skip search initialization. */
/* Load KWS if one was specified in config */
if ((keyphrase = cmd_ln_str_r(config, "-keyphrase"))) {
if (ps_set_keyphrase(ps, PS_DEFAULT_SEARCH, keyphrase))
return -1;
ps_set_search(ps, PS_DEFAULT_SEARCH);
}
if ((path = cmd_ln_str_r(config, "-kws"))) {
if (ps_set_kws(ps, PS_DEFAULT_SEARCH, path))
return -1;
ps_set_search(ps, PS_DEFAULT_SEARCH);
}
/* Load an FSG if one was specified in config */
if ((path = cmd_ln_str_r(config, "-fsg"))) {
fsg_model_t *fsg = fsg_model_readfile(path, ps->lmath, lw);
if (!fsg)
return -1;
if (ps_set_fsg(ps, PS_DEFAULT_SEARCH, fsg))
return -1;
ps_set_search(ps, PS_DEFAULT_SEARCH);
}
/* Or load a JSGF grammar */
if ((path = cmd_ln_str_r(config, "-jsgf"))) {
if (ps_set_jsgf_file(ps, PS_DEFAULT_SEARCH, path)
|| ps_set_search(ps, PS_DEFAULT_SEARCH))
return -1;
}
if ((path = cmd_ln_str_r(ps->config, "-allphone"))) {
if (ps_set_allphone_file(ps, PS_DEFAULT_SEARCH, path)
|| ps_set_search(ps, PS_DEFAULT_SEARCH))
return -1;
}
if ((path = cmd_ln_str_r(ps->config, "-lm")) &&
!cmd_ln_boolean_r(ps->config, "-allphone")) {
if (ps_set_lm_file(ps, PS_DEFAULT_SEARCH, path)
|| ps_set_search(ps, PS_DEFAULT_SEARCH))
return -1;
}
if ((path = cmd_ln_str_r(ps->config, "-lmctl"))) {
const char *name;
ngram_model_t *lmset;
ngram_model_set_iter_t *lmset_it;
if (!(lmset = ngram_model_set_read(ps->config, path, ps->lmath))) {
E_ERROR("Failed to read language model control file: %s\n", path);
return -1;
}
for(lmset_it = ngram_model_set_iter(lmset);
lmset_it; lmset_it = ngram_model_set_iter_next(lmset_it)) {
ngram_model_t *lm = ngram_model_set_iter_model(lmset_it, &name);
E_INFO("adding search %s\n", name);
if (ps_set_lm(ps, name, lm)) {
ngram_model_free(lm);
ngram_model_set_iter_free(lmset_it);
return -1;
}
ngram_model_free(lm);
}
name = cmd_ln_str_r(config, "-lmname");
if (name)
ps_set_search(ps, name);
else {
E_ERROR("No default LM name (-lmname) for `-lmctl'\n");
return -1;
}
}
/* Initialize performance timer. */
ps->perf.name = "decode";
ptmr_init(&ps->perf);
return 0;
}
