#include <lm_state.h>

Collaboration diagram for tesseract::LanguageModelNgramInfo:

Public Member Functions
	LanguageModelNgramInfo (const char *c, int l, bool p, float nc, float ncc)

Public Attributes
STRING	context

int	context_unichar_step_len

bool	pruned

float	ngram_cost
	-ln(P_ngram_model(path)) More...

float	ngram_and_classifier_cost
	-[ ln(P_classifier(path)) + scale_factor * ln(P_ngram_model(path)) ] More...

Detailed Description

Struct for storing additional information used by Ngram language model component.

Constructor & Destructor Documentation

◆ LanguageModelNgramInfo()

tesseract::LanguageModelNgramInfo::LanguageModelNgramInfo	(	const char *	c,
		int	l,
		bool	p,
		float	nc,
		float	ncc
	)

inline

Member Data Documentation

◆ context

STRING tesseract::LanguageModelNgramInfo::context

◆ context_unichar_step_len

int tesseract::LanguageModelNgramInfo::context_unichar_step_len

Length of the context measured by advancing using UNICHAR::utf8_step() (should be at most the order of the character ngram model used).

◆ ngram_and_classifier_cost

float tesseract::LanguageModelNgramInfo::ngram_and_classifier_cost

-[ ln(P_classifier(path)) + scale_factor * ln(P_ngram_model(path)) ]

◆ ngram_cost

float tesseract::LanguageModelNgramInfo::ngram_cost

-ln(P_ngram_model(path))

◆ pruned

bool tesseract::LanguageModelNgramInfo::pruned

The paths with pruned set are pruned out from the perspective of the character ngram model. They are explored further because they represent a dictionary match or a top choice. Thus ngram_info is still computed for them in order to calculate the combined cost.

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

/home/stephane/src/tesseract/src/wordrec/lm_state.h

Public Member Functions

Public Attributes