構造体 WCHMM_INFO

Whole lexicon tree structure holding all information. [詳細]

#include <wchmm.h>

WCHMM_INFOのコラボレーション図

Collaboration graph
[凡例]

変数

int lmtype
 LM type
int lmvar
 LM variant
boolean category_tree
 TRUE if category_tree is used
HTK_HMM_INFOhmminfo
 HMM definitions used to construct this lexicon
NGRAM_INFOngram
 N-gram used to construct this lexicon
DFA_INFOdfa
 Grammar used to construct this lexicon
WORD_INFOwinfo
 Word dictionary used to construct this lexicon
boolean ccd_flag
 TRUE if handling context dependency
int maxwcn
 Memory assigned maximum number of nodes
int n
 Num of nodes in this lexicon
WCHMM_STATEstate
 HMM state on tree lexicon [nodeID]
LOGPROBself_a
 Transition probability to self node
LOGPROBnext_a
 Transition probabiltiy to next (now+1) node
A_CELL2 ** ac
 Transition arc information other than self and next.
WORD_IDstend
 Word ID that ends at the state [nodeID]
int ** offset
 Node ID of a phone [wordID][0..phonelen-1]
int * wordend
 Node ID of word-end state [wordID]
int startnum
 Number of root nodes
int * startnode
 Root node index [0..startnum-1] -> node ID
int * wordbegin
 Node ID of word-beginning state [wordID] for multipath mode
int maxstartnum
 Allocated number of startnodes for multipath mode
WORD_IDstart2wid
 Root node index [0..startnum-1] -> word ID for multipath mode
LOGPROBwordend_a
 Transition prob. outside word [wordID] for non-multipath mode
S_CELL ** sclist
 List of successor list [scid]
int * sclist2node
 Mapping successor list [scid] to node
int scnum
 Number of factoring nodes that has successor list
BMALLOC_BASEmalloc_root
 Pointer for block memory allocation
HMMWorkhmmwrk
 Work area for HMM computation in wchmm
LM_PROB_CACHE lmcache
 LM score cache for 1st pass
WCHMM_WORK wrk
 Other work area for 1st pass transition computation
int separated_word_count
 Number of words actually separated (linearlized) from the tree
LOGPROB(* uni_prob_user )(WORD_INFO *, WORD_ID, LOGPROB)
 Pointer to function returning word occurence probability
LOGPROB(* bi_prob_user )(WORD_INFO *, WORD_ID, WORD_ID, LOGPROB)
 Pointer to function returning a word probability given a word context (corresponds to bi-gram)

説明

Whole lexicon tree structure holding all information.

wchmm.h215 行で定義されています。


この構造体の説明は次のファイルから生成されました:
Juliusに対してTue Dec 18 16:05:28 2007に生成されました。  doxygen 1.5.4