class LMNGram

This class is the main class for N-gram language modelling.

Public Fields

int n

The order of the N-Gram

real* unigrams

The unigram probability entries, 1 for every vocab word

LMInteriorLevelWordList* next_level

Points to the start of the language model tree structure

LMCache* cache

The cache used for fast LM lookup

Public Methods

LMNGram( int n_, Vocabulary* vocab_ )

Creates an empty N-Gram data structure.

Public Members

Adds a new entry to the N-gram

'order' is the order of the entry (ie.

The 'words' array needs to be in oldest-word-first order

([w1 w2 w3] for trigram entry)

'prob' is the log probability for the entry

'bo_wt' is the log back-off weight for the entry.

Finds the N-gram probability of a given word sequence, with full

backoff

'order' is the number of words in the 'words' array

The ordering in 'words' is W3,W2,W1,W4 for a 4-gram query

Documentation

This class is the main class for N-gram language modelling. After language model entries have been added, lookups with full backoff can be performed.

int n

The order of the N-Gram

real* unigrams

The unigram probability entries, 1 for every vocab word

LMInteriorLevelWordList* next_level

Points to the start of the language model tree structure

LMCache* cache

The cache used for fast LM lookup

LMNGram( int n_, Vocabulary* vocab_ )

Creates an empty N-Gram data structure. 'n_' is the N-gram order.

Adds a new entry to the N-gram

Adds a new entry to the N-gram

'order' is the order of the entry (ie.

'order' is the order of the entry (ie. 2 for a bigram entry).

The 'words' array needs to be in oldest-word-first order

The 'words' array needs to be in oldest-word-first order

([w1 w2 w3] for trigram entry)

([w1 w2 w3] for trigram entry)

'prob' is the log probability for the entry

'prob' is the log probability for the entry

'bo_wt' is the log back-off weight for the entry.

'bo_wt' is the log back-off weight for the entry. If order is

Finds the N-gram probability of a given word sequence, with full

Finds the N-gram probability of a given word sequence, with full

backoff

backoff

'order' is the number of words in the 'words' array

'order' is the number of words in the 'words' array

The ordering in 'words' is W3,W2,W1,W4 for a 4-gram query

The ordering in 'words' is W3,W2,W1,W4 for a 4-gram query

This class has no child classes.

Author:

Darren Moore (moore@idiap.ch)

Alphabetic index HTML hierarchy of classes or Java