Julius: libjulius/src/ngram_decode.c File Reference

These functions returns next word candidates in the 2nd recognition pass of Julius, i.e. N-gram based stack decoding.

Given a partial sentence hypothesis, it first estimate the beginning frame of the hypothesis based on the word trellis. Then the words in the word trellis around the estimated frame are extracted from the word trellis. They will be returned with their N-gram probabilities.

Function Documentation

static int compare_nw	(	NEXTWORD **	a,
		NEXTWORD **	b
	)			`[static]`

qsort callback function to sort next word candidates by their word ID.

Parameters:

	a	[in] element 1
	b	[in] element 2

Returns:: 1 if word id of a > that of b, -1 if negative, 0 if equal.

Definition at line 69 of file ngram_decode.c.

Referenced by get_backtrellis_words().

static NEXTWORD* search_nw	(	NEXTWORD **	nw,
		WORD_ID	w,
		int	num
	)			`[static]`

Find a word from list of next word candidates.

Parameters:

	nw	[in] list of next word candidates
	w	[in] word id to search for
	num	[in] length of nw

Returns:: the pointer to the NEXTWORD data if found, or NULL if not found.

Definition at line 99 of file ngram_decode.c.

static LOGPROB ngram_forw2back	(	NGRAM_INFO *	ngram,
		WORD_ID *	w,
		int	wlen
	)			`[static]`

Compute backward N-gram score from forward N-gram.

Parameters:

	ngram	[in] N-gram data structure
	w	[in] word sequence
	wlen	[in] length of w

Returns:: the backward probability of the word w[0].

Definition at line 139 of file ngram_decode.c.

static int pick_backtrellis_words	(	RecogProcess *	r,
		NEXTWORD **	nw,
		int	oldnum,
		NODE *	hypo,
		short	t
	)			`[static]`

Extract next word candidates from word trellis.

This function extracts the list of trellis words whose word end has survived in the word trellis at the specified frame. The N-gram probabilities of them are then computed and added to the current next word candidates data.

Parameters:

	r	[in] recognition process instance
	nw	[in] list of next word candidates (new words will be appended at oldnum)
	oldnum	[in] number of words already stored in nw
	hypo	[in] the source sentence hypothesis
	t	[in] specified frame

Returns:: the total number of words currently stored in the nw.

< Last two non-transparent words

< Num of found non-transparent words (<=2)

< Num of skipped transparent words

Definition at line 192 of file ngram_decode.c.

Referenced by get_backtrellis_words().

static int get_backtrellis_words	(	RecogProcess *	r,
		NEXTWORD **	nw,
		NODE *	hypo,
		short	tm,
		short	t_end
	)			`[static]`

Determine next word candidates from the word trellis.

This function builds a list of next word candidates by looking up the word trellis at specified frame, with lookup_range frame margin. If the same words exists in the near frames, only the one nearest to the specified frame will be chosen.

Parameters:

	r	[in] recognition process instance
	nw	[out] pointer to hold the extracted words as list of next word candidates
	hypo	[in] partial sentence hypothesis from which the words will be expanded
	tm	[in] center time frame to look up the words
	t_end	[in] right frame boundary for the lookup.

Returns:: the number of next words candidates stored in nw.

Definition at line 334 of file ngram_decode.c.

Referenced by ngram_nextwords().

static int limit_nw	(	NEXTWORD **	nw,
		NODE *	hypo,
		int	num,
		WORD_INFO *	winfo
	)			`[static]`

Remove non-expansion word from list.

Remove words in the nextword list which should not be expanded.

Parameters:

	nw	[i/o] list of next word candidates (will be shrinked by removing some words)
	hypo	[in] partial sentence hypothesis from which the words will be expanded
	num	[in] current number of next words in nw
	winfo	[in] word dictionary

Returns:: the new number of words in nw

Definition at line 428 of file ngram_decode.c.

Referenced by ngram_nextwords().

int ngram_firstwords	(	NEXTWORD **	nw,
		int	peseqlen,
		int	maxnw,
		RecogProcess *	r
	)

Get initial word hypotheses at the beginning.

on N-gram based recogntion, the initial hypothesis is fixed to the tail silence word. Exception is that, in short-pause segmentation mode, the initial hypothesis will be chosen from survived words on the last input frame in the first pass.

Parameters:

	nw	[out] pointer to hold the initial word candidates
	peseqlen	[in] input frame length
	maxnw	[in] maximum number of words that can be stored in nw
	r	[in] recognition process instance

Returns:: the number of words extracted and stored to nw.

Definition at line 496 of file ngram_decode.c.

Referenced by wchmm_fbs().

Here is the caller graph for this function:

int ngram_nextwords	(	NODE *	hypo,
		NEXTWORD **	nw,
		int	maxnw,
		RecogProcess *	r
	)

Return the list of next word candidate.

Given a partial sentence hypothesis "hypo", it returns the list of next word candidates. Actually, it extracts from word trellis the list of words whose word-end node has survived near the estimated beginning-of-word frame of last word "hypo->estimated_next_t", and store them to "nw" with their N-gram probabilities.

Parameters:

	hypo	[in] source partial sentence hypothesis
	nw	[out] pointer to store the list of next word candidates (should be already allocated)
	maxnw	[in] maximum number of words that can be stored to nw
	r	[in] recognition process instance

Returns:: the number of extracted next word candidates in nw.

Definition at line 563 of file ngram_decode.c.

boolean ngram_acceptable	(	NODE *	hypo,
		RecogProcess *	r
	)

Acceptance check.

Return whether the given partial hypothesis is acceptable as a sentence and can be treated as a final search candidate. In N-gram mode, it checks whether the last word is the beginning-of-sentence silence (silhead).

Parameters:

	hypo	[in] partial sentence hypothesis to be examined
	r	[in] recognition process instance

Returns:: TRUE if acceptable as a sentence, or FALSE if not.

Definition at line 613 of file ngram_decode.c.

libjulius/src/ngram_decode.c File Reference

Functions

Detailed Description

Function Documentation


Functions
static int	compare_nw (NEXTWORD a, NEXTWORD b)
	qsort callback function to sort next word candidates by their word ID.
static NEXTWORD *	search_nw (NEXTWORD **nw, WORD_ID w, int num)
	Find a word from list of next word candidates.
static LOGPROB	ngram_forw2back (NGRAM_INFO ngram, WORD_ID w, int wlen)
	Compute backward N-gram score from forward N-gram.
static int	pick_backtrellis_words (RecogProcess r, NEXTWORD nw, int oldnum, NODE hypo, short t)
	Extract next word candidates from word trellis.
static int	get_backtrellis_words (RecogProcess r, NEXTWORD nw, NODE hypo, short tm, short t_end)
	Determine next word candidates from the word trellis.
static int	limit_nw (NEXTWORD *nw, NODE hypo, int num, WORD_INFO *winfo)
	Remove non-expansion word from list.
int	ngram_firstwords (NEXTWORD *nw, int peseqlen, int maxnw, RecogProcess r)
	Get initial word hypotheses at the beginning.
int	ngram_nextwords (NODE hypo, NEXTWORD nw, int maxnw, RecogProcess r)
	Return the list of next word candidate.
boolean	ngram_acceptable (NODE hypo, RecogProcess r)
	Acceptance check.