search_bestfirst_main.c

来自「julius version 4.12.about sound recognit」· C语言 代码 · 共 2,134 行 · 第 1/5 页

    } else if (r->lmtype == LM_DFA) {
      acc = dfa_acceptable(now, r);
    }
    if (acc && now->estimated_next_t <= 5) {
      new = newnode(r);
      /* new に now の面咳をコピ〖して·呵姜弄なスコアを纷换 */
      /* copy content of 'now' to 'new', and compute the final score */
      last_next_word(now, new, param, r);
      if (debug2_flag) {
	jlog("DEBUG:  This is acceptable as a sentence candidate\n");
      }
      /* g[] が掐蜗幌眉に茫していなければ逮笛 */
      /* reject this sentence candidate if g[] does not reach the end */
      if (new->score <= LOG_ZERO) {
	if (debug2_flag) {
	  jlog("DEBUG:  But invalid because Viterbi pass does not reach the 0th frame\n");
	}
	free_node(new);
	free_node(now);
	continue;
      }
      /* 减妄フラグを惟てて掐れ木す */
      /* set endflag and push again  */
      if (debug2_flag) {
	jlog("DEBUG  This hypo itself was pushed with final score=%f\n", new->score);
      }
      new->endflag = TRUE;
      if (put_to_stack(new, &start, &bottom, &stacknum, stacksize) != -1) {
	if (r->graphout) {
	  if (new->score > LOG_ZERO) {
	    new->lastcontext = now->prevgraph;
	    new->prevgraph = wordgraph_assign(new->seq[new->seqnum-1],
					      WORD_INVALID,
					      (new->seqnum >= 2) ? new->seq[new->seqnum-2] : WORD_INVALID,
					      0,
#ifdef GRAPHOUT_PRECISE_BOUNDARY
					      /* wordend are shifted to the last */
#ifdef PASS2_STRICT_IWCD
					      new->wordend_frame[0],
#else
					      now->wordend_frame[0],
#endif
#else
					      now->bestt,
#endif
					      new->score,
					      prev_score,
					      now->g[0],
#ifdef GRAPHOUT_PRECISE_BOUNDARY
#ifdef PASS2_STRICT_IWCD
					      new->wordend_gscore[0],
#else
					      now->wordend_gscore[0],
#endif
#else
					      now->tail_g_score,
#endif
					      now->lscore,
#ifdef CM_SEARCH
					      new->cmscore[new->seqnum-1],
#else
					      LOG_ZERO,
#endif
					      r
					      );
	  } else {
	    new->lastcontext = now->lastcontext;
	    new->prevgraph = now->prevgraph;
	  }
	} /* put_to_stack() != -1 */
      }	/* recog->graphout */
      /* この簿棱はここで姜わらずに, ここからさらに帽胳鸥倡する */
      /* continue with the 'now' hypothesis, not terminate here */
    }
    
    /*
     * この簿棱から·肌帽胳礁圭を疯年する. 
     * 肌帽胳礁圭は, この簿棱の夸年幌眉フレ〖ム件收に赂哼した
     * 妈１パスのトレリス帽胳礁圭. 
     *
     * N-gramの眷圭は称帽胳の n-gram 儡鲁澄唯が崔まれる. 
     * DFA の眷圭は, その面でさらに DFA 惧で儡鲁材墙なもののみが手ってくる
     */
    /*
     * Determine next word set that can connect to this hypothesis.
     * They come from the trellis word that has been survived at near the
     * beginning of the last word.
     *
     * In N-gram mode, they also contain N-gram probabilities toward the
     * source hypothesis.  In DFA mode, the word set is further reduced
     * by the grammatical constraint
     */
#ifdef DEBUG
    jlog("DEBUG: get next words\n");
#endif
    if (r->lmtype == LM_PROB) {
      nwnum = ngram_nextwords(now, nextword, maxnwnum, r);
    } else if (r->lmtype == LM_DFA) {
      nwnum = dfa_nextwords(now, nextword, maxnwnum, r);
      /* nextword が邦れたら、バッファを笼やして浩チャレンジ */
      /* If the number of nextwords can exceed the buffer size, expand the
	 nextword data area */
      while (nwnum < 0) {
	nextword = nw_expand(nextword, &maxnwnum, &nwroot, winfo->num);
	nwnum = dfa_nextwords(now, nextword, maxnwnum, r);
      }
    }
    if (debug2_flag) {
      jlog("DEBUG:  %d words extracted from wordtrellis\n", nwnum);
    }

    /* 
     * 簿棱と肌帽胳礁圭から糠たな矢簿棱を栏喇し·スタックにいれる. 
     */
    /*
     * generate new hypotheses from 'now' and 'nextword', 
     * and push them to stack
     */
#ifdef DEBUG
    jlog("DEBUG: generate hypo\n");
#endif

    if (r->lmtype == LM_DFA) {
      now_noise_calced = FALSE;	/* TRUE is noise-inserted score has been calculated */
    }
    i = dwrk->pushctr;		/* store old value */

#ifdef CM_SEARCH
    /* initialize local stack */
    cm_init(dwrk, winfo->num, jconf->annotate.cm_alpha
#ifdef CM_MULTIPLE_ALPHA
	    , jconf->annotate.cm_alpha_num
#endif
	    );
#endif

    /* for each nextword, generate a new hypothesis */
    for (w = 0; w < nwnum; w++) {
      if (r->lmtype == LM_DFA) {
	/* limit word hypothesis */
	if (! (winfo->wton[nextword[w]->id] >= cate_bgn && winfo->wton[nextword[w]->id] < cate_bgn + cate_num)) {
	  continue;
	}
      }
      new = newnode(r);

      if (r->lmtype == LM_DFA) {

	if (nextword[w]->can_insert_sp == TRUE) {
	  /* ノイズを洞んだトレリススコアを纷换し·洞まない眷圭との呵络猛を艰る */
	  /* compute hypothesis score with noise inserted */
	  
	  if (now_noise_calced == FALSE) {
	    /* now に sp をつけた簿棱 now_noise を侯り,そのスコアを纷换 */
	    /* generate temporal hypothesis 'now_noise' which has short-pause
	       word after the original 'now' */
	    fornoise.id = gdfa->sp_id;
	    now_noise = newnode(r);
	    cpy_node(now_noise, now);
#if 0
	    now_noise_tmp = newnode(r);
	    next_word(now, now_noise_tmp, &fornoise, param, r);
	    scan_word(now_noise_tmp, param, r);
	    for(t=0;t<peseqlen;t++) {
	      now_noise->g[t] = max(now_noise_tmp->g[t], now->g[t]);
	    }
	    free_node(now_noise_tmp);
#else
	    /* expand NOISE only if it exists in backward trellis */
	    /* begin patch by kashima */
	    if (jconf->pass2.looktrellis_flag) {
	      if(!dfa_look_around(&fornoise, now, r)){
		free_node(now_noise);
		free_node(new);
		continue;
	      }
	    }
	    /* end patch by kashima */
	    
	    /* now_nosie の スコア g[] を纷换し·傅の now の g[] と孺秤して
	       光い数を何脱 */
	    /* compute trellis score g[], and adopt the maximum score
	       for each frame compared with now->g[] */
	    next_word(now, now_noise, &fornoise, param, r);
	    scan_word(now_noise, param, r);
	    for(t=0;t<peseqlen;t++) {
	      now_noise->g[t] = max(now_noise->g[t], now->g[t]);
	    }
	    /* ノイズを洞んだ狠を雇胃したスコアを纷换したので·
	       ここで呵稿のノイズ帽胳を now_noise から久す */
	    /* now that score has been computed considering pause insertion,
	       we can delete the last noise word from now_noise here */
	    now_noise->seqnum--;
#endif
	    now_noise_calced = TRUE;
	  }
	  
	  /* expand word only if it exists in backward trellis */
	  /* begin patch by kashima */
	  if (jconf->pass2.looktrellis_flag) {
	    if(!dfa_look_around(nextword[w], now_noise, r)){
	      free_node(new);
	      continue;
	    }
	  }
	  /* end patch by kashima */
	  
	  /* 糠しい簿棱' new' を 'now_noise' から栏喇 */
	  /* generate a new hypothesis 'new' from 'now_noise' */
	  next_word(now_noise, new, nextword[w], param, r);
	  
	} else {
	  
	  /* expand word only if it exists in backward trellis */
	  /* begin patch by kashima */
	  if (jconf->pass2.looktrellis_flag) {
	    if(!dfa_look_around(nextword[w], now, r)){
	      free_node(new);
	      continue;
	    }
	  }
	  /* end patch by kashima */
	  
	  /* 糠しい簿棱' new' を 'now_noise' から栏喇 */
	  /* generate a new hypothesis 'new' from 'now_noise' */
	  next_word(now, new, nextword[w], param, r);
	  
	}
      }

      if (r->lmtype == LM_PROB) {

	/* 糠しい簿棱' new' を 'now_noise' から栏喇
	   N-gram の眷圭はノイズを泼侍胺いしない */
	/* generate a new hypothesis 'new' from 'now'.
	   pause insertion is treated as same as normal words in N-gram mode. */
	next_word(now, new, nextword[w], param, r);

      }

      if (new->score <= LOG_ZERO) { /* not on trellis */
	free_node(new);
	continue;
      }
	
      dwrk->genectr++;

#ifdef CM_SEARCH
      /* store the local hypothesis to temporal stack */
      cm_store(dwrk, new);
#else 
      /* 栏喇した簿棱 'new' をスタックに掐れる */
      /* push the generated hypothesis 'new' to stack */

      /* stack overflow */
      if (can_put_to_stack(new, &bottom, &stacknum, stacksize) == -1) {
	free_node(new);
	continue;
      }
      
      if (r->graphout) {
	/* assign a word arc to the last fixed word */
	new->lastcontext = now->prevgraph;
	new->prevgraph = wordgraph_assign(new->seq[new->seqnum-2],
					  new->seq[new->seqnum-1],
					  (new->seqnum >= 3) ? new->seq[new->seqnum-3] : WORD_INVALID,
					  new->bestt + 1,
#ifdef GRAPHOUT_PRECISE_BOUNDARY
#ifdef PASS2_STRICT_IWCD
					  /* most up-to-date wordend_gscore is on new, because the last phone of 'now' will be computed at next_word() */
					  new->wordend_frame[new->bestt],
#else
					  now->wordend_frame[new->bestt],
#endif
#else
					  now->bestt,
#endif
					  new->score, prev_score,
#ifdef PASS2_STRICT_IWCD
					  new->g[new->bestt] - new->lscore,
#else
					  now->g[new->bestt+1],
#endif
#ifdef GRAPHOUT_PRECISE_BOUNDARY
#ifdef PASS2_STRICT_IWCD
					  /* most up-to-date wordend_gscore is on new, because the last phone of 'now' will be computed at next_word() */
					  new->wordend_gscore[new->bestt],
#else
					  now->wordend_gscore[new->bestt],
#endif
#else
					  now->tail_g_score,
#endif
					  now->lscore,
#ifdef CM_SEARCH
					  new->cmscore[new->seqnum-2],
#else
					  LOG_ZERO,
#endif
					  r
					  );
      }	/* recog->graphout */
      put_to_stack(new, &start, &bottom, &stacknum, stacksize);
      if (debug2_flag) {
	j = new->seq[new->seqnum-1];
	jlog("DEBUG:  %15s [%15s](id=%5d)(%f) [%d-%d] pushed\n",winfo->wname[j], winfo->woutput[j], j, new->score, new->estimated_next_t + 1, new->bestt);
      }
      dwrk->current = new;
      //callback_exec(CALLBACK_DEBUG_PASS2_PUSH, r);
      dwrk->pushctr++;
#endif
    } /* end of nextword loop */

#ifdef CM_SEARCH
    /* compute score sum */
    cm_sum_score(dwrk
#ifdef CM_MULTIPLE_ALPHA
		 , jconf->annotate.cm_alpha_bgn 
		 , jconf->annotate.cm_alpha_end
		 , jconf->annotate.cm_alpha_step
#endif
		 );
    /* compute CM and put the generated hypotheses to global stack */
    while ((new = cm_get_node(dwrk)) != NULL) {
      cm_set_score(dwrk, new
#ifdef CM_MULTIPLE_ALPHA
		   , jconf->annotate.cm_alpha_bgn
		   , jconf->annotate.cm_alpha_end
		   , jconf->annotate.cm_alpha_step
#endif
		   );
#ifdef CM_SEARCH_LIMIT
      if (new->cmscore[new->seqnum-1] < jconf->annotate.cm_cut_thres
#ifdef CM_SEARCH_LIMIT_AFTER
	  && dwrk->finishnum > 0
#endif
	  ) {
	free_node(new);
	continue;
      }
#endif /* CM_SEARCH_LIMIT */
      /*      j = new->seq[new->seqnum-1];
	      printf("  %15s [%15s](id=%5d)(%f) [%d-%d] cm=%f\n",winfo->wname[j], winfo->woutput[j], j, new->score, new->estimated_next_t + 1, new->bestt, new->cmscore[new->seqnum-1]);*/

      /* stack overflow */
      if (can_put_to_stack(new, &bottom, &stacknum, stacksize) == -1) {
	free_node(new);
	continue;
      }

      if (r->graphout) {

	/* assign a word arc to the last fixed word */
	new->lastcontext = now->prevgraph;
	new->prevgraph = wordgraph_assign(new->seq[new->seqnum-2],
					  new->seq[new->seqnum-1],
					  (new->seqnum >= 3) ? new->seq[new->seqnum-3] : WORD_INVALID,
					  new->bestt + 1,
#ifdef GRAPHOUT_PRECISE_BOUNDARY
#ifdef PASS2_STRICT_IWCD
					  new->wordend_frame[new->bestt],
#else
					  now->wordend_frame[new->bestt],
#endif
#else
					  now->bestt,
#endif
					  new->score, prev_score,
#ifdef PASS2_STRICT_IWCD
					  new->g[new->bestt] - new->lscore,
#else
					  now->g[new->bestt+1],
#endif
#ifdef GRAPHOUT_PRECISE_BOUNDARY
#ifdef PASS2_STRICT_IWCD
					  new->wordend_gscore[new->bestt],
#else
					  now->wordend_gscore[new->bestt],
#endif
#else
					  now->tail_g_score,
#endif
					  now->lscore,
#ifdef CM_SEARCH
					  new->cmscore[new->seqnum-2],
#else
					  LOG_ZERO,
#endif
					  r
					  );
      }	/* recog->graphout */
      
      put_to_stack(new, &start, &bottom, &stacknum, stacksize);
      if (debug2_flag) {
	j = new->seq[new->seqnum-1];
	jlog("DEBUG:  %15s [%15s](id=%5d)(%f) [%d-%d] pushed\n",winfo->wname[j], winfo->woutput[j], j, new->score, new->estimated_next_t + 1, new->bestt);
      }
      dwrk->current = new;
      //callback_exec(CALLBACK_DEBUG_PASS2_PUSH, r);
      dwrk->pushctr++;
    }
#endif

    if (debug2_flag) {
      jlog("DEBUG: %d pushed\n",dwrk->pushctr-i);
    }
    if (r->lmtype == LM_DFA) {
      if (now_noise_calced == TRUE) free_node(now_noise);
    }
    
    /* 
     * 艰り叫した簿棱を嘉てる
     * free the source hypothesis
     */
    free_node(now);

  }
  /***************/
  /* End of Loop */
  /***************/

  /* output */
  if (dwrk->finishnum == 0) {		/* if search failed */

    /* finalize result when no hypothesis was obtained */
    if (verbo