libsent/include/sent/vocabulary.h File Reference

Word dictionary for recognition. More...

#include <sent/stddefs.h>
#include <sent/htk_hmm.h>

Go to the source code of this file.

Data Structures

struct  WORD_INFO
 Word dictionary structure to hold vocabulary. More...

Defines

#define BEGIN_WORD_DEFAULT   "<s>"
 Default word string of beginning-of-sentence word.
#define END_WORD_DEFAULT   "</s>"
 Default word string of end-of-sentence word.
#define MAXWSTEP   4000
 Memory allocation step in number of words when loading a word dictionary.

Functions

WORD_INFO * word_info_new ()
 Allocate a new word dictionary structure.
void word_info_free (WORD_INFO *winfo)
 Free all informations in the WORD_INFO.
void winfo_init (WORD_INFO *winfo)
 Initialize a new word dictionary structure.
boolean winfo_expand (WORD_INFO *winfo)
 Expand the word dictionary.
boolean init_voca (WORD_INFO *winfo, char *filename, HTK_HMM_INFO *hmminfo, boolean, boolean)
 Load and initialize a word dictionary.
boolean init_wordlist (WORD_INFO *winfo, char *filename, HTK_HMM_INFO *hmminfo, char *headphone, char *tailphone, char *contextphone, boolean force_dict)
 Load and initialize a word list for isolated word recognition.
void voca_set_stats (WORD_INFO *winfo)
 Parse a word dictionary and set the maximum state length per word.
void voca_load_start (WORD_INFO *winfo, HTK_HMM_INFO *hmminfo, boolean ignore_tri_conv)
 Start loading a dictionary.
boolean voca_load_line (char *buf, WORD_INFO *winfo, HTK_HMM_INFO *hmminfo)
 Load a line from buffer and set parameters to the dictionary.
boolean voca_load_end (WORD_INFO *winfo)
 End loading dictionary entries.
boolean voca_load_htkdict (FILE *, WORD_INFO *, HTK_HMM_INFO *, boolean)
 Top function to read word dictionary via file pointer.
boolean voca_load_htkdict_fd (int, WORD_INFO *, HTK_HMM_INFO *, boolean)
 Top function to read word dictionary via file descriptor.
boolean voca_load_htkdict_sd (int, WORD_INFO *, HTK_HMM_INFO *, boolean)
 Top function to read word dictionary via socket descriptor.
boolean voca_append_htkdict (char *entry, WORD_INFO *winfo, HTK_HMM_INFO *hmminfo, boolean ignore_tri_conv)
 Append a single entry to the existing word dictionary.
boolean voca_append (WORD_INFO *dstinfo, WORD_INFO *srcinfo, int coffset, int woffset)
 Append one word dictionary to other, for multiple grammar handling.
boolean voca_load_htkdict_line (char *buf, WORD_ID *vnum, int linenum, WORD_INFO *winfo, HTK_HMM_INFO *hmminfo, boolean do_conv, boolean *ok_flag)
 Sub function to Add a dictionary entry line to the word dictionary.
boolean voca_load_word_line (char *buf, WORD_INFO *winfo, HTK_HMM_INFO *hmminfo, char *headphone, char *tailpohone, char *contextphone)
 Load a line from buffer and set parameters to the dictionary.
boolean voca_load_wordlist (FILE *fp, WORD_INFO *winfo, HTK_HMM_INFO *hmminfo, char *headphone, char *tailphone, char *contextphone)
 Top function to read word list via file pointer.
boolean voca_load_wordlist_line (char *buf, WORD_ID *vnum, int linenum, WORD_INFO *winfo, HTK_HMM_INFO *hmminfo, boolean do_conv, boolean *ok_flag, char *headphone, char *tailphone, char *contextphone)
 Sub function to Add a dictionary entry line to the word dictionary.
boolean voca_mono2tri (WORD_INFO *winfo, HTK_HMM_INFO *hmminfo)
 Convert whole words in word dictionary to word-internal triphone.
WORD_ID voca_lookup_wid (char *, WORD_INFO *)
 Look up a word on dictionary by string.
WORD_ID * new_str2wordseq (WORD_INFO *, char *, int *)
 Convert string of space-separated word strings to array of word ids.
char * cycle_triphone (char *p)
 Return string of triphone name composed from last 3 call.
char * cycle_triphone_flush ()
 Flush the triphone buffer and return the last biphone.
void print_voca_info (FILE *fp, WORD_INFO *)
 Output overall word dictionary information to stdout.
void put_voca (FILE *fp, WORD_INFO *winfo, WORD_ID wid)
 Output information of a word in dictionary to stdout.
boolean make_base_phone (HTK_HMM_INFO *hmminfo, WORD_INFO *winfo)
 Build basephone information.
void print_phone_info (FILE *fp, HTK_HMM_INFO *hmminfo)
 Output general information concerning phone mapping in HMM definition.
void print_all_basephone_detail (HMM_basephone *base)
 Output all basephone informations to stdout.
void print_all_basephone_name (HMM_basephone *base)
 Output all basephone names to stdout.
void test_interword_triphone (HTK_HMM_INFO *hmminfo, WORD_INFO *winfo)
 Top function to check if all the possible triphones on given word dictionary actually exist in the logical HMM.

Detailed Description

Word dictionary for recognition.

This file defines data structure for word dictionary used in recognition. It stores word's string, output string, phoneme sequence, transparency. Beginning-of-sentence word and End-of-sentence word guessed from runtime environment is also stored here.

Please note that the N-gram vocabulary is stored in NGRAM_INFO and it can differ from this word dictionary. The reference from the word dictionary to a N-gram vocabulary is done by wton[] member in WORD_INFO. When used with DFA, the wton[] holds a category number to which each word belongs.

Author:
Akinobu LEE
Date:
Sat Feb 12 12:38:13 2005
Revision
1.1.1.1

Definition in file vocabulary.h.

Function Documentation

WORD_INFO* word_info_new (  ) 

Allocate a new word dictionary structure.

Returns:
pointer to the newly allocated WORD_INFO.

Definition at line 35 of file voca_malloc.c.

Referenced by initialize_dict(), multigram_read_file_and_add(), multigram_update(), and read_grammar_from_socket().

void word_info_free ( WORD_INFO *  winfo  ) 

Free all informations in the WORD_INFO.

Parameters:
winfo [i/o] word dictionary data to be freed.

Definition at line 51 of file voca_malloc.c.

Referenced by initialize_dict(), j_process_lm_free(), multigram_exec_delete(), multigram_free_all(), multigram_read_file_and_add(), and multigram_update().

void winfo_init ( WORD_INFO *  winfo  ) 

Initialize a new word dictionary structure.

Parameters:
winfo [i/o] word dictionary to be initialized.

Definition at line 75 of file voca_malloc.c.

Referenced by multigram_update(), and voca_load_start().

boolean winfo_expand ( WORD_INFO *  winfo  ) 

Expand the word dictionary.

Parameters:
winfo [i/o] word dictionary to be expanded.

Definition at line 105 of file voca_malloc.c.

Referenced by voca_load_line(), and voca_load_word_line().

boolean init_voca ( WORD_INFO *  winfo,
char *  filename,
HTK_HMM_INFO *  hmminfo,
boolean  not_conv_tri,
boolean  force_dict 
)

Load and initialize a word dictionary.

Parameters:
winfo [out] pointer to a word dictionary data to store the read data
filename [in] file name of the word dictionary to read
hmminfo [in] HMM definition data, needed for triphone conversion.
not_conv_tri [in] TRUE if not converting monophone to triphone.
force_dict [in] TRUE if want to ignore the error words in the dictionary
Returns:
TRUE on success, FALSE on failure.

Definition at line 41 of file init_voca.c.

Referenced by initialize_dict(), and multigram_read_file_and_add().

boolean init_wordlist ( WORD_INFO *  winfo,
char *  filename,
HTK_HMM_INFO *  hmminfo,
char *  headphone,
char *  tailphone,
char *  contextphone,
boolean  force_dict 
)

Load and initialize a word list for isolated word recognition.

Parameters:
winfo [out] pointer to a word dictionary data to store the read data
filename [in] file name of the word dictionary to read
hmminfo [in] HMM definition data, needed for triphone conversion.
headphone [in] word head silence phone name
tailphone [in] word tail silence phone name
conextphone [in] silence context name at head and tail phoneme
force_dict [in] TRUE if want to ignore the error words in the dictionary
Returns:
TRUE on success, FALSE on failure.

Definition at line 81 of file init_voca.c.

Referenced by multigram_read_file_and_add().

void voca_set_stats ( WORD_INFO *  winfo  ) 

Parse a word dictionary and set the maximum state length per word.

Parameters:
winfo [i/o]

Definition at line 186 of file voca_load_htkdict.c.

Referenced by voca_load_end().

void voca_load_start ( WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
boolean  ignore_tri_conv 
)

Start loading a dictionary.

See voca_load_htkdict() for an example of using this function.

Parameters:
winfo [i/o] dictionary data where the data will be loaded
hmminfo [in] phoneme HMM definition
ignore_tri_conv [in] if TRUE, skip triphone conversion while loading

Definition at line 228 of file voca_load_htkdict.c.

Referenced by voca_load_htkdict(), voca_load_htkdict_fd(), voca_load_htkdict_sd(), and voca_load_wordlist().

boolean voca_load_line ( char *  buf,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo 
)

Load a line from buffer and set parameters to the dictionary.

See voca_load_htkdict() for an example of using this function.

Parameters:
buf [in] input buffer containing a word entry
winfo [i/o] word dictionary to append the entry
hmminfo [in] phoneme HMM definition
Returns:
TRUE when successfully read, or FALSE on encountered end of dictionary. When an error occurs, this function will set winfo->ok_flag to FALSE.

Definition at line 255 of file voca_load_htkdict.c.

Referenced by voca_append_htkdict(), voca_load_htkdict(), voca_load_htkdict_fd(), and voca_load_htkdict_sd().

boolean voca_load_end ( WORD_INFO *  winfo  ) 

End loading dictionary entries.

It calculates some statistics for the read entries, outputs errors if encountered while the last loading, and returns with status whether an error occured while loading.

Parameters:
winfo [i/o] word dictionary just read by voca_load_line() calls
Returns:
TRUE when no error has been occured during loading, or FALSE if an error occured.

Definition at line 284 of file voca_load_htkdict.c.

Referenced by voca_append_htkdict(), voca_load_htkdict(), voca_load_htkdict_fd(), voca_load_htkdict_sd(), and voca_load_wordlist().

boolean voca_load_htkdict ( FILE *  fp,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
boolean  ignore_tri_conv 
)

Top function to read word dictionary via file pointer.

Parameters:
fp [in] file pointer
winfo [out] pointer to word dictionary to store the read data.
hmminfo [in] HTK HMM definition data. if NULL, phonemes are ignored.
ignore_tri_conv [in] TRUE if triphone conversion is ignored
Returns:
TRUE on success, FALSE on any error word.

Definition at line 305 of file voca_load_htkdict.c.

Referenced by init_voca().

boolean voca_load_htkdict_fd ( int  fd,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
boolean  ignore_tri_conv 
)

Top function to read word dictionary via file descriptor.

Parameters:
fd [in] file descriptor
winfo [out] pointer to word dictionary to store the read data.
hmminfo [in] HTK HMM definition data. if NULL, phonemes are ignored.
ignore_tri_conv [in] TRUE if triphone conversion is ignored
Returns:
TRUE on success, FALSE on any error word.

Definition at line 330 of file voca_load_htkdict.c.

Referenced by read_grammar_from_socket().

boolean voca_load_htkdict_sd ( int  sd,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
boolean  ignore_tri_conv 
)

Top function to read word dictionary via socket descriptor.

Parameters:
sd [in] socket descriptor
winfo [out] pointer to word dictionary to store the read data.
hmminfo [in] HTK HMM definition data. if NULL, phonemes are ignored.
ignore_tri_conv [in] TRUE if triphone conversion is ignored
Returns:
TRUE on success, FALSE on any error word.

Definition at line 354 of file voca_load_htkdict.c.

Referenced by read_grammar_from_socket().

boolean voca_append_htkdict ( char *  entry,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
boolean  ignore_tri_conv 
)

Append a single entry to the existing word dictionary.

Parameters:
entry [in] dictionary entry string to be appended.
winfo [out] pointer to word dictionary to append the data.
hmminfo [in] HTK HMM definition data. if NULL, phonemes are ignored.
ignore_tri_conv [in] TRUE if triphone conversion is ignored
Returns:
TRUE on success, FALSE on any error word.

Definition at line 378 of file voca_load_htkdict.c.

Referenced by initialize_dict().

boolean voca_append ( WORD_INFO *  dstinfo,
WORD_INFO *  srcinfo,
int  coffset,
int  woffset 
)

Append one word dictionary to other, for multiple grammar handling.

Assumes that the same HMM definition is used on both word dictionary.

Parameters:
dstinfo [i/o] word dictionary
srcinfo [in] word dictionary to be appended to dst
coffset [in] category id offset in dst where the new data should be stored
woffset [in] word id offset in dst where the new data should be stored

Definition at line 664 of file voca_load_htkdict.c.

Referenced by multigram_add_words_to_grammar(), multigram_append_to_global(), and multigram_update().

boolean voca_load_htkdict_line ( char *  buf,
WORD_ID *  vnum_p,
int  linenum,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
boolean  do_conv,
boolean *  ok_flag 
)

Sub function to Add a dictionary entry line to the word dictionary.

Parameters:
buf [i/o] buffer to hold the input string, will be modified in this function
vnum_p [in] current number of words in winfo
linenum [in] current line number of the input
winfo [out] pointer to word dictionary to append the data.
hmminfo [in] HTK HMM definition data. if NULL, phonemes are ignored.
do_conv [in] TRUE if performing triphone conversion
ok_flag [out] will be set to FALSE if an error occured for this input.
Returns:
FALSE if buf == "DICEND", else TRUE will be returned.

Definition at line 398 of file voca_load_htkdict.c.

Referenced by voca_load_line().

boolean voca_load_word_line ( char *  buf,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
char *  headphone,
char *  tailphone,
char *  contextphone 
)

Load a line from buffer and set parameters to the dictionary.

Parameters:
buf [in] input buffer containing a word entry
winfo [i/o] word dictionary to append the entry
hmminfo [in] phoneme HMM definition
headphone [in] word head silence model name
tailphone [in] word tail silence model name
contextphone [in] silence context name to be used at head and tail
Returns:
TRUE when successfully read, or FALSE on encountered end of dictionary. When an error occurs, this function will set winfo->ok_flag to FALSE.

Definition at line 114 of file voca_load_wordlist.c.

Referenced by voca_load_wordlist().

boolean voca_load_wordlist ( FILE *  fp,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
char *  headphone,
char *  tailphone,
char *  contextphone 
)

Top function to read word list via file pointer.

Parameters:
fp [in] file pointer
winfo [out] pointer to word dictionary to store the read data.
hmminfo [in] HTK HMM definition data. if NULL, phonemes are ignored.
headphone [in] word head silence model name
tailphone [in] word tail silence model name
contextphone [in] silence context name to be used at head and tail
Returns:
TRUE on success, FALSE on any error word.

Definition at line 142 of file voca_load_wordlist.c.

Referenced by init_wordlist().

boolean voca_load_wordlist_line ( char *  buf,
WORD_ID *  vnum_p,
int  linenum,
WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo,
boolean  do_conv,
boolean *  ok_flag,
char *  headphone,
char *  tailphone,
char *  contextphone 
)

Sub function to Add a dictionary entry line to the word dictionary.

Parameters:
buf [i/o] buffer to hold the input string, will be modified in this function
vnum_p [in] current number of words in winfo
linenum [in] current line number of the input
winfo [out] pointer to word dictionary to append the data.
hmminfo [in] HTK HMM definition data. if NULL, phonemes are ignored.
do_conv [in] TRUE if performing triphone conversion
ok_flag [out] will be set to FALSE if an error occured for this input.
headphone [in] word head silence model name
tailphone [in] word tail silence model name
contextphone [in] silence context name to be used at head and tail
Returns:
FALSE if buf == "DICEND", else TRUE will be returned.

Definition at line 172 of file voca_load_wordlist.c.

Referenced by voca_load_word_line().

boolean voca_mono2tri ( WORD_INFO *  winfo,
HTK_HMM_INFO *  hmminfo 
)

Convert whole words in word dictionary to word-internal triphone.

Normally triphone conversion will be performed directly when reading dictionary file. This function is for post conversion only.

Parameters:
winfo [i/o] word dictionary information
hmminfo [in] HTK HMM definition
Returns:
TRUE on success, FALSE on failure.

Definition at line 623 of file voca_load_htkdict.c.

WORD_ID voca_lookup_wid ( char *  keyword,
WORD_INFO *  winfo 
)

Look up a word on dictionary by string.

Parameters:
keyword [in] keyword to search
winfo [in] word dictionary
Returns:
the word id if found, or WORD_INVALID if not found.

Definition at line 43 of file voca_lookup.c.

Referenced by initialize_dict(), and new_str2wordseq().

WORD_ID* new_str2wordseq ( WORD_INFO *  winfo,
char *  s,
int *  len_return 
)

Convert string of space-separated word strings to array of word ids.

Parameters:
winfo [in] word dictionary
s [in] string of space-separated word strings
len_return [out] number of found words
Returns:
pointer to a newly allocated word list.

Definition at line 117 of file voca_lookup.c.

char* cycle_triphone ( char *  p  ) 

Return string of triphone name composed from last 3 call.

Parameters:
p [in] next phone string
Returns:
the composed triphone name, or NULL on end.

Definition at line 80 of file voca_load_htkdict.c.

Referenced by cycle_triphone_flush(), new_str2phseq(), voca_load_htkdict_line(), and voca_load_wordlist_line().

char* cycle_triphone_flush (  ) 

Flush the triphone buffer and return the last biphone.

Returns:
the composed last bi-phone name.

Definition at line 126 of file voca_load_htkdict.c.

Referenced by new_str2phseq(), voca_load_htkdict_line(), and voca_load_wordlist_line().

void print_voca_info ( FILE *  fp,
WORD_INFO *  winfo 
)

Output overall word dictionary information to stdout.

Parameters:
fp [in] file descriptor
winfo [in] word dictionary

Definition at line 35 of file voca_util.c.

Referenced by print_engine_info().

void put_voca ( FILE *  fp,
WORD_INFO *  winfo,
WORD_ID  wid 
)

Output information of a word in dictionary to stdout.

Parameters:
fp [in] file descriptor
winfo [in] word dictionary
wid [in] word id to be output

Definition at line 67 of file voca_util.c.

Referenced by hmm_check(), make_dfa_voca_ref(), print_engine_info(), and wchmm_add_word().

boolean make_base_phone ( HTK_HMM_INFO *  hmminfo,
WORD_INFO *  winfo 
)

Build basephone information.

Extract base phones from HMM definition, mark them whether they appear on word head or word tail, and count the number.

Parameters:
hmminfo [i/o] HMM definition information, basephone list will be added.
winfo [in] word dictionary information

Definition at line 386 of file chkhmmlist.c.

Referenced by hmm_check().

void print_phone_info ( FILE *  fp,
HTK_HMM_INFO *  hmminfo 
)

Output general information concerning phone mapping in HMM definition.

Parameters:
fp [in] file descriptor
hmminfo [in] HMM definition data.

Definition at line 404 of file chkhmmlist.c.

Referenced by hmm_check().

void print_all_basephone_detail ( HMM_basephone *  base  ) 

Output all basephone informations to stdout.

Parameters:
base [in] pointer to the top basephone data holder.

Definition at line 106 of file chkhmmlist.c.

Referenced by hmm_check().

void print_all_basephone_name ( HMM_basephone *  base  ) 

Output all basephone names to stdout.

Parameters:
base [in] pointer to the top basephone data holder.

Definition at line 116 of file chkhmmlist.c.

Referenced by hmm_check().

void test_interword_triphone ( HTK_HMM_INFO *  hmminfo,
WORD_INFO *  winfo 
)

Top function to check if all the possible triphones on given word dictionary actually exist in the logical HMM.

Parameters:
hmminfo [in] HMM definition information, with basephone list.
winfo [in] word dictionary information

Definition at line 345 of file chkhmmlist.c.

Referenced by hmm_check().

Generated on Tue Dec 18 16:01:37 2007 for Julius by  doxygen 1.5.4