NGRAM_TUPLE_INFO Struct Reference

N-gram entries for a m-gram (1 <= m <= N). More...

#include <ngram2.h>

Data Fields
int	totalnum
	Number of defined tuples.
boolean	is24bit
	TRUE if this m-gram uses 24bit index for tuples instead of 32bit.
int	bgnlistlen
	Length of bgn and num, should be the same as context_num of (m-1)-gram.
NNID_UPPER *	bgn_upper
	Beginning ID of a tuple set whose context is the (m-1) tuple for 24bit mode (upper 8bit).
NNID_LOWER *	bgn_lower
	Beginning ID of a tuple set whose context is the (m-1) tuple for 24bit mode (lower 16bit).
NNID *	bgn
	Beginning ID of a tuple set whose context is the (m-1) tuple for 32bit mode.
WORD_ID *	num
	Size of a tuple set whose context is the (m-1) tuple.
WORD_ID *	nnid2wid
	List of Word IDs of edge word of the tuple.
LOGPROB *	prob
	Log probabilities of edge word of the tuple.
int	context_num
	Number of tuples to be a context of (m+1)-gram (= number of defined back-off weights).
LOGPROB *	bo_wt
	Back-off weights for (m+1)-gram, the length is context_num if ct_compaction is TRUE, or totalnum if FALSE.
boolean	ct_compaction
	TRUE if use compacted index for back-off contexts.
NNID_UPPER *	nnid2ctid_upper
	Index to map tuple ID of this m-gram to valid context id (upper 8bit).
NNID_LOWER *	nnid2ctid_lower
	Index to map tuple ID of this m-gram to valid context id (upper 16bit).

---

Detailed Description

N-gram entries for a m-gram (1 <= m <= N).

Definition at line 129 of file ngram2.h.

---

The documentation for this struct was generated from the following file:

• libsent/include/sent/ngram2.h

---