hdman.c

Hidden Markov Toolkit (HTK) 3.2.1 HTK is a toolkit for use in research into automatic speech recogn

   db->nextWord=labels[0];
   db->pbuf.source=db->source;
   if (labels[1]!=NULL) 
      db->nextOutSym=labels[1];
   else 
      db->nextOutSym=labels[0];
   db->pbuf.phone[0]=wdBnd;
   for (i=0;i<n;i++)
      db->pbuf.phone[i+1]=labels[i+2];
   db->pbuf.phone[n+1]=wdBnd;
   db->pbuf.nPhone=n+2;
   db->pbuf.prob=prob;

   return TRUE;
}

/* WriteEntry: write out a dictionary entry and update newPhones */
void WriteEntry(FILE *f, LabId word, LabId outsym, Pronunciation *p, int margin, Boolean findNew)
{
   int i,st,en;
   char buf[256],m[20];
   
   if (p->nPhone == 0) return;
   if (!nullOutput) {
      if (word != NULL)
         strcpy(buf,ReWriteString(word->name,NULL,ESCAPE_CHAR));
      else
         strcpy(buf,"");
      for (i=0; i<margin; i++) m[i] = ' '; m[i]='\0';
      fprintf(f,"%s%-15s",m,buf);
      if (incOutSyms) {
         strcpy(buf,"["); 
         strcat(buf,ReWriteString(outsym->name,NULL,ESCAPE_CHAR)); 
         strcat(buf,"]");
         fprintf(f," %-15s",buf);
      }
      if (incProbs) {
         if (p->prob<1.0)
            fprintf(f," %8.6f",p->prob);
         else
            fprintf(f,"         ");
      }
   }
   st = 1;
   if  (p->phone[0] != wdBnd){
      st = 0;
      if (isLogging && margin==0)
         fprintf(logF,"WARNING: no left word bnd in word %s\n",word->name);
   }
   en = p->nPhone-1;
   if (p->phone[en] != wdBnd){
      ++en;
      if (isLogging && margin==0)
         fprintf(logF,"WARNING: no right word bnd in word %s\n",word->name);
   }      
   for (i=st; i<en; i++) { 
      if (!nullOutput)
         fprintf(f," %s",ReWriteString((p->phone[i])->name,NULL,ESCAPE_CHAR));
      if (findNew)
         PutPhone(p->phone[i]); 
   }
   if (tagSources && !nullOutput)
      fprintf(f,"   [%s]",ReWriteString(p->source->name,NULL,ESCAPE_CHAR));
   if (!nullOutput) fprintf(f,"\n");
}

/* ----------------- Read/Write Dictionary Words --------------- */

/* ReadDictProns: read entries for next word, return FALSE if none found */
Boolean ReadDictProns(DBuffer *db)
{
   LabId thisWord;
   Boolean ok;
   int i = 0;

   if (db->nextWord == NULL) return FALSE;
   if (db->wbuf.word != NULL && strcmp(db->wbuf.word->name,db->nextWord->name) > 0 )
      HError(1452,"ReadDictProns: word %s out of order in dict %s",
             db->nextWord->name,db->name);
   db->wbuf.word = thisWord = db->nextWord;
   db->wbuf.outsym = db->nextOutSym;
   db->wbuf.pron[i++]=db->pbuf;
   ok = ReadNextWord(db);
   while (ok && db->nextWord == thisWord){
      if (i == MAXPRONS)
         HError(1430,"ReadDictProns: max prons exceeded in word %s from dict %s",
                thisWord->name,db->name);
      db->wbuf.pron[i++]=db->pbuf;
      ok = ReadNextWord(db);
   }
   db->wbuf.nPron = i;
   ++db->totalWords;
   db->totalProns += i;
   return TRUE;
}

/* WriteDictWord: write current word to output file */
void WriteDictWord(DBuffer *db, FILE *f, int margin, Boolean findNew)
{
   int i;

   if (f==NULL) return;
   for (i=0; i<db->wbuf.nPron; i++)
      WriteEntry(f,db->wbuf.word,db->wbuf.outsym,db->wbuf.pron+i,margin, findNew);
}

void ShowDB(DBuffer *db, char * title)
{
   printf("%s - DB %s: word %s, next %s\n",
          title,
          db->name,
          db->wbuf.word==NULL?"NULL":db->wbuf.word->name,
          db->nextWord==NULL ?"NULL":db->nextWord->name);
   fflush(stdout);
}

/* --------------------------- Editing ------------------------- */

/* IsInIdList: return true if id is in idlist */
Boolean IsInIdList(LabId id, LabId *idlist)
{
   while (*idlist != NULL) {
      if (id == *idlist) return TRUE;
      ++idlist;
   }
   return FALSE;
}

/* UCPhoneOp: Upper Case Phone command */
void UCPhoneOp(WordBuf *wb)
{
   int i,j;
   Pronunciation *p;
   
   for (i=0;i<wb->nPron; i++){
      p = wb->pron+i;
      for (j=0; j<p->nPhone; j++)
         p->phone[j] = UCase(p->phone[j]);
   }
}

/* LCPhoneOp: Lower Case Phone command */
void LCPhoneOp(WordBuf *wb)
{
   int i,j;
   Pronunciation *p;
   
   for (i=0; i<wb->nPron; i++){
      p = wb->pron+i;
      for (j=0; j<p->nPhone; j++)
         p->phone[j] = LCase(p->phone[j]);
   }
}

/* DeleteWordOp: Delete Word Command - marks by setting nPron to zero */
void DeleteWordOp(WordBuf *wb, LabId *args)
{
   if (IsInIdList(wb->word,args))
      wb->nPron = 0;
}

/* DeleteSourceOp: del source if more than one */
int DeleteSourceOp(WordBuf *wb, LabId *args)
{
   int i,j,k,del=0;
   Pronunciation *p, *s, *t;
   
   if (wb->nPron < 2) return 0;
   for (i=0; i<wb->nPron; i++){
      p = wb->pron+i;
      if (p->source == *args){
         if (trace&T_DSOP)
            printf("Removing pron %s from word %s\n",
                   (*args)->name,wb->word->name);
         --wb->nPron;
         ++del;
         for (j=i; j<wb->nPron; j++){
            t = wb->pron+j; s = t+1;
            t->nPhone = s->nPhone;
            t->source = s->source;
            t->prob = s->prob;
            for (k=0; k<t->nPhone; k++)
               t->phone[k] = s->phone[k];
         }
      }
   }
   return del;
}

/* DelDefOp: Delete Word Command - marks by setting nPron to zero */
void DelDefOp(WordBuf *wb, LabId *args)
{
   int i,j,idx=0;
   Boolean found = FALSE;
   Pronunciation *p;
   LabId *ph;
   char buf[256];

   if (wb->word == *args){
      for (i=0; !found && i<wb->nPron; i++){
         found = TRUE; idx = i; p = wb->pron+i; ph = args+1;
         for (j=1; *ph != NULL && j<p->nPhone; j++,ph++){
            if (p->phone[j] != *ph){
               found = FALSE; break;
            }
         }
      }
      if (found){
         if (trace&T_DWOP){
            buf[0] = '\0';
            for (ph=args; *ph!=NULL; ph++){
               strcat(buf," "); strcat(buf,(*ph)->name);
            }
            printf("      deleting definition %s\n",buf);
         }
         for (j=idx+1; j<wb->nPron; j++)
            wb->pron[j-1] = wb->pron[j];
         --wb->nPron;
      }else{
         buf[0] = '\0';
         for (ph=args; *ph!=NULL; ph++){
            strcat(buf," "); strcat(buf,(*ph)->name);
         }
         HError(-1431,"DelDefOp: pron %s not found",buf);
      }
   }
}

/* FunctionWordOp: Function Word Command - makes fn word phones of form W.A */
void FunctionWordOp(WordBuf *wb, LabId *args)
{
   int i,j;
   static char s[255];
   Pronunciation *p;
   
   if (IsInIdList(wb->word,args)) 
      for (i=0; i<wb->nPron; i++){
         p = wb->pron+i;
         for (j=1; j<p->nPhone-1; j++){
            sprintf(s,"%s.%s",(wb->word)->name,(p->phone[j])->name);
            p->phone[j] = GetLabId(s,TRUE);
         }
      }
}  

/* ReplacePhoneOp: Replace Phone Command */
void ReplacePhoneOp(WordBuf *wb, LabId *args)
{
   int i,j;
   Pronunciation *p;
   
   for (i=0; i<wb->nPron; i++){
      p = wb->pron+i;
      for (j=0; j<p->nPhone; j++)
         if (IsInIdList(p->phone[j],args+1)) 
            p->phone[j] = *args;
   }
} 

/* ReplaceWordOp: Replace Word Command */
void ReplaceWordOp(WordBuf *wb, LabId *args)
{
   if (IsInIdList(wb->word,args+1)) {
      if (wb->word == wb->outsym)
         wb->word = wb->outsym = *args;
      else
         wb->word = *args;
   }
} 

/* GetContextList: extract context list for id if any */   
LabId *GetContextList(LabId id, DBuffer *db)
{
   int i;
   LabId *list;

   for (i=0; i<db->numCons; i++){
      list = db->contexts[i].args;
      if (id == *list) return list+1;
   }
   return NULL;
}

/* ContextRep: do context replace on given pronunciation - note 
   * matches any context */
void ContextRep(Pronunciation *p, LabId *args, DBuffer *db)
{
   LabId lc,rc,cc;
   Boolean ltrue, replace;
   LabId *lcList,*rcList;
   int i;
   
   lc = *(args+1); cc = *(args+2); rc = *(args+3);
   lcList = GetContextList(lc,db);
   rcList = GetContextList(rc,db);
   for (i=1; i<p->nPhone-1; i++) {
      replace = FALSE;
      if (p->phone[i] == cc) {
         if (lcList != NULL)
            ltrue = IsInIdList(p->phone[i-1],lcList);
         else
            ltrue = (lc == asterix || lc == p->phone[i-1] );
         if (ltrue){
            if (rcList != NULL)
               replace = IsInIdList(p->phone[i+1],rcList);
            else
               replace = (rc == asterix || rc == p->phone[i+1] );
         }
      }
      if (replace) p->phone[i] = *args;
   }  
}

/* ContextReplaceOp: ContextReplace Command */
void ContextReplaceOp(WordBuf *wb, LabId *args, DBuffer *db)
{
   int i;
   Pronunciation *p;
   
   for (i=0; i<wb->nPron; i++){
      p = wb->pron+i;
      ContextRep(p,args,db);
   }
}        

/* SeqMatch: Returns true if the nMerge sequence of LabIds 
             are the same in list1 and list2 */
Boolean SeqMatch(int nMerge, LabId *list1, LabId *list2)
{
   int i;

   for (i=0; i<nMerge; i++,list1++,list2++)
      if (*list1 != *list2)
         return FALSE;
   return TRUE;
}
   
/* MergePhon: merge all occs of phone sequence in args */
void MergePhon(Pronunciation *p, int nArgs, LabId *args)
{
   int i,j,nMerge;
   
   nMerge = nArgs-1;
   for (i=0;i<=p->nPhone-nMerge;i++)
      if (SeqMatch(nMerge,p->phone+i,args+1)) {
         p->phone[i] = *args;
         for (j=i+1;j<p->nPhone;j++)
            p->phone[j] = p->phone[j+nMerge-1];
         p->nPhone -= nMerge-1;    
      }
}


/* MergePhoneOp: MergePhone Command */
void MergePhoneOp(WordBuf *wb, int nArgs, LabId *args)
{
   int i;
   Pronunciation *p;
   
   for (i=0; i<wb->nPron; i++){
      p = wb->pron+i;
      MergePhon(p,nArgs,args);
   }
}        

/* SplitPhon: split a phone sequence into multiple phones */
void SplitPhon(Pronunciation *p, int nArgs, LabId *args)
{
   int i,j,nSplit,nExtra;
   
   nSplit = nArgs-1;
   nExtra = nSplit-1;
   for (i=0;i<p->nPhone;)
      if (p->phone[i] == args[0]) {
         if (p->nPhone+nExtra > MAXPHONES)
            HError(1430,"SplitPhon: Too many phones to split");
         for (j=p->nPhone-1+nExtra; j>=i+1; j--)
            p->phone[j] = p->phone[j-nExtra];
         for (j=0; j<nSplit; j++)
            p->phone[i+j] = args[j+1];
         i += nExtra+1;
         p->nPhone += nExtra;
      }
      else i++;
}            

/* SplitPhoneOp: SplitPhone Command */
void SplitPhoneOp(WordBuf *wb, int nArgs, LabId *args)
{
   int i;
   Pronunciation *p;
   
   for (i=0; i<wb->nPron; i++){
      p = wb->pron+i;
      SplitPhon(p,nArgs,args);
   }
}        

/* DeletePhoneOp: DeletePhone Command */
void DeletePhoneOp(WordBuf *wb, LabId *args)
{
   int i,j,k;
   Pronunciation *p;
   
   for (k=0; k<wb->nPron; k++){
      p = wb->pron+k;
      for (i=0;i<p->nPhone;)
         if (IsInIdList(p->phone[i],args)) {
            for (j=i+1;j<p->nPhone;j++)
               p->phone[j-1] = p->phone[j]; 
            p->nPhone--;
         }
         else
            i++;
   }
}

/* AppendPhone: append id to given pronunciation */
void AppendPhone(Pronunciation *p, LabId id)
{
   if (p->nPhone == MAXPHONES)
      HError(1430,"AppendPhone: MAXPHONES exceeded");
   p->phone[p->nPhone-1] = id;
   p->phone[p->nPhone++] = wdBnd;
}

/* DuplicatePron: duplicate selected pron and insert at end of wbuf */
void DuplicatePron(WordBuf *wb, int i)
{
   int j;
   Pronunciation *s, *t;

   if (wb->nPron == MAXPRONS)
      HError(1430,"DuplicatePron: MAXPRONS exceeded");
   s = wb->pron+i;
   t = wb->pron+wb->nPron;
   t->source = s->source;
   t->nPhone = s->nPhone;
   t->prob = s->prob;
   for (j=0; j<s->nPhone; j++) t->phone[j] = s->phone[j];
   ++wb->nPron;   
}

/* AppendSilenceOp: AppendSilences Command */
void AppendSilenceOp(WordBuf *wb, LabId *args)
{
   LabId *i;
   int j,n,k;

   n = wb->nPron-1;
   for (j=n; j>=0; j--){
      for (i=args+1; *i != NULL; i++){
         DuplicatePron(wb,j);
         k = wb->nPron-1;
         AppendPhone(wb->pron+k,*i);
      }
      AppendPhone(wb->pron+j, *args);
   }
}

/* MakeTriId:  concatenate args separated by - and +'s and return its id */
LabId MakeTriId(LabId l, LabId c, LabId r)
{
   char buf[100];