libsent/include/sent/vocabulary.h

単語辞書の構造体定義 [詳細]

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

ソースコードを見る。

データ構造

struct  WORD_INFO
 Word dictionary structure to hold vocabulary [詳細]

マクロ定義

#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

関数

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.

説明

単語辞書の構造体定義

このファイルは認識で用いられる単語辞書を定義します．単語辞書は単語の読み， 出力文字列，音素列の他に，文の開始・終了単語や透過単語情報も保持します．

N-gramに出現する語彙の辞書は NGRAM_INFO に格納され，この認識用単語辞書とは 区別されることに注意して下さい．単語辞書からN-gramの語彙へのマッピングは WORD_INFO 内の wton[] によって行なわれます．またDFAの場合，wton は その単語が属するDFA_INFO 内のカテゴリ番号を含みます．

作者:
Akinobu LEE
日付:
Sat Feb 12 12:38:13 2005
Revision
1.1.1.1

vocabulary.h で定義されています。

関数

WORD_INFO* word_info_new (  ) 

Allocate a new word dictionary structure.

戻り値:
pointer to the newly allocated WORD_INFO.

voca_malloc.c の 35 行で定義されています。

参照元 initialize_dict(), multigram_read_file_and_add(), multigram_update(), と read_grammar_from_socket().

void word_info_free ( WORD_INFO *  winfo  ) 

Free all informations in the WORD_INFO.

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

voca_malloc.c の 51 行で定義されています。

参照元 initialize_dict(), j_process_lm_free(), multigram_exec_delete(), multigram_free_all(), multigram_read_file_and_add(), と multigram_update().

void winfo_init ( WORD_INFO *  winfo  ) 

Initialize a new word dictionary structure.

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

voca_malloc.c の 75 行で定義されています。

参照元 multigram_update(), と voca_load_start().

boolean winfo_expand ( WORD_INFO *  winfo  ) 

Expand the word dictionary.

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

voca_malloc.c の 105 行で定義されています。

参照元 voca_load_line(), と 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.

引数:
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
戻り値:
TRUE on success, FALSE on failure.

init_voca.c の 41 行で定義されています。

参照元 initialize_dict(), と 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.

引数:
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
戻り値:
TRUE on success, FALSE on failure.

init_voca.c の 81 行で定義されています。

参照元 multigram_read_file_and_add().

void voca_set_stats ( WORD_INFO *  winfo  ) 

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

引数:
winfo [i/o]

voca_load_htkdict.c の 186 行で定義されています。

参照元 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.

引数:
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

voca_load_htkdict.c の 228 行で定義されています。

参照元 voca_load_htkdict(), voca_load_htkdict_fd(), voca_load_htkdict_sd(), と 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.

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

voca_load_htkdict.c の 255 行で定義されています。

参照元 voca_append_htkdict(), voca_load_htkdict(), voca_load_htkdict_fd(), と 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.

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

voca_load_htkdict.c の 284 行で定義されています。

参照元 voca_append_htkdict(), voca_load_htkdict(), voca_load_htkdict_fd(), voca_load_htkdict_sd(), と 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

引数:
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
戻り値:
TRUE on success, FALSE on any error word.

voca_load_htkdict.c の 305 行で定義されています。

参照元 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.

引数:
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
戻り値:
TRUE on success, FALSE on any error word.

voca_load_htkdict.c の 330 行で定義されています。

参照元 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.

引数:
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
戻り値:
TRUE on success, FALSE on any error word.

voca_load_htkdict.c の 354 行で定義されています。

参照元 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.

引数:
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
戻り値:
TRUE on success, FALSE on any error word.

voca_load_htkdict.c の 378 行で定義されています。

参照元 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.

引数:
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

voca_load_htkdict.c の 664 行で定義されています。

参照元 multigram_add_words_to_grammar(), multigram_append_to_global(), と 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.

引数:
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.
戻り値:
FALSE if buf == "DICEND", else TRUE will be returned.

voca_load_htkdict.c の 398 行で定義されています。

参照元 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.

引数:
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
戻り値:
TRUE when successfully read, or FALSE on encountered end of dictionary. When an error occurs, this function will set winfo->ok_flag to FALSE.

voca_load_wordlist.c の 114 行で定義されています。

参照元 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

引数:
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
戻り値:
TRUE on success, FALSE on any error word.

voca_load_wordlist.c の 142 行で定義されています。

参照元 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.

引数:
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
戻り値:
FALSE if buf == "DICEND", else TRUE will be returned.

voca_load_wordlist.c の 172 行で定義されています。

参照元 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.

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

voca_load_htkdict.c の 623 行で定義されています。

WORD_ID voca_lookup_wid ( char *  keyword,
WORD_INFO *  winfo 
)

Look up a word on dictionary by string.

引数:
keyword [in] keyword to search
winfo [in] word dictionary
戻り値:
the word id if found, or WORD_INVALID if not found.

voca_lookup.c の 43 行で定義されています。

参照元 initialize_dict(), と 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.

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

voca_lookup.c の 117 行で定義されています。

char* cycle_triphone ( char *  p  ) 

Return string of triphone name composed from last 3 call.

引数:
p [in] next phone string
戻り値:
the composed triphone name, or NULL on end.

voca_load_htkdict.c の 80 行で定義されています。

参照元 cycle_triphone_flush(), new_str2phseq(), voca_load_htkdict_line(), と voca_load_wordlist_line().

char* cycle_triphone_flush (  ) 

Flush the triphone buffer and return the last biphone.

戻り値:
the composed last bi-phone name.

voca_load_htkdict.c の 126 行で定義されています。

参照元 new_str2phseq(), voca_load_htkdict_line(), と voca_load_wordlist_line().

void print_voca_info ( FILE *  fp,
WORD_INFO *  winfo 
)

Output overall word dictionary information to stdout.

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

voca_util.c の 35 行で定義されています。

参照元 print_engine_info().

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

Output information of a word in dictionary to stdout.

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

voca_util.c の 67 行で定義されています。

参照元 hmm_check(), make_dfa_voca_ref(), print_engine_info(), と 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.

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

chkhmmlist.c の 386 行で定義されています。

参照元 hmm_check().

void print_phone_info ( FILE *  fp,
HTK_HMM_INFO *  hmminfo 
)

Output general information concerning phone mapping in HMM definition.

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

chkhmmlist.c の 404 行で定義されています。

参照元 hmm_check().

void print_all_basephone_detail ( HMM_basephone *  base  ) 

Output all basephone informations to stdout.

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

chkhmmlist.c の 106 行で定義されています。

参照元 hmm_check().

void print_all_basephone_name ( HMM_basephone *  base  ) 

Output all basephone names to stdout

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

chkhmmlist.c の 116 行で定義されています。

参照元 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.

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

chkhmmlist.c の 345 行で定義されています。

参照元 hmm_check().

Juliusに対してTue Dec 18 16:05:03 2007に生成されました。  doxygen 1.5.4