lti::hmmOnlineClassifier Class Reference

This class allows online classification of word chains. More...

#include <ltiHmmOnlineClassifier.h>

Inheritance diagram for lti::hmmOnlineClassifier:

[legend]

Collaboration diagram for lti::hmmOnlineClassifier:

[legend]

List of all members.

Classes
class	parameters
	the parameters for the class hmmOnlineClassifier More...
Public Member Functions
	hmmOnlineClassifier ()
	hmmOnlineClassifier (const hmmOnlineClassifier &other)
virtual	~hmmOnlineClassifier ()
virtual const char *	getTypeName () const
hmmOnlineClassifier &	copy (const hmmOnlineClassifier &other)
hmmOnlineClassifier &	operator= (const hmmOnlineClassifier &other)
virtual classifier *	clone () const
const parameters &	getParameters () const
virtual bool	write (ioHandler &handler, const bool complete=true) const
virtual bool	read (ioHandler &handler, const bool complete=true)
Online classification routines.
These methods are used for online classification.
void	reset ()
void	initialize ()
bool	classify (const dvector &observation)
trace back routines
These methods are used to determine the best word sequence.
ivector	partialTraceBack ()
ivector	partialTraceBack (ivector &wordLength, dvector &wordEndScores)
ivector	fullTraceBack () const
ivector	fullTraceBack (ivector &wordLength, dvector &wordEndScores) const
ivector	bestHypothesisTraceBack () const
ivector	bestHypothesisTraceBack (int &innerState, ivector &wordLength, dvector &wordEndScores) const
int	getNumberOfHypotheses () const
word error routines
These methods are used to determine the word error rate of a classified sentence.
int	wordErrorCount (const ivector &wordSequence1, const ivector &wordSequence2) const
float	wordErrorRate (const ivector &correctSequence, const ivector &classifiedSequence) const

---

Detailed Description

This class allows online classification of word chains.

---

Constructor & Destructor Documentation

lti::hmmOnlineClassifier::hmmOnlineClassifier

(

)

default constructor

lti::hmmOnlineClassifier::hmmOnlineClassifier

(

const hmmOnlineClassifier &

other

)

copy constructor

Parameters:

other

the object to be copied

virtual lti::hmmOnlineClassifier::~hmmOnlineClassifier

(

)

[virtual]

destructor

---

Member Function Documentation

ivector lti::hmmOnlineClassifier::bestHypothesisTraceBack	(	int &	innerState,
		ivector &	wordLength,
		dvector &	wordEndScores
	)			const

Full traceback of the best active hypothesis.

Return the word sequence leading to the currently best hypotheses. Note, that this hypothesis might not be in a word-end-state.

Parameters:

	innerState	the active state of the hypothesis is saved here. the according model is given by the last entry of the return vector.
	wordLength	the length of each classified word is saved here.
	wordEndScores	the score of the word end hypotheses is saved here.

Returns:

vector describing a sequence of word ids.

ivector lti::hmmOnlineClassifier::bestHypothesisTraceBack

(

)

const

Full traceback of the best active hypothesis.

Return the word sequence leading to the currently best hypotheses. Note, that this hypothesis might not be in a word-end-state.

Returns:

vector describing a sequence of word ids.

bool lti::hmmOnlineClassifier::classify

(

const dvector &

observation

)

Online classification step.

Parameters:

observation

the observation of this timestep.

Returns:

false if an error occurred during classification else true

virtual classifier* lti::hmmOnlineClassifier::clone

(

)

const [virtual]

returns a pointer to a clone of this functor.

Reimplemented from lti::hmmClassifier.

hmmOnlineClassifier& lti::hmmOnlineClassifier::copy

(

const hmmOnlineClassifier &

other

)

copy data of "other" functor.

Parameters:

other

the functor to be copied

Returns:

a reference to this functor object

Reimplemented from lti::hmmClassifier.

ivector lti::hmmOnlineClassifier::fullTraceBack	(	ivector &	wordLength,
		dvector &	wordEndScores
	)			const

Full word sequence traceback Return the best word sequence, given the assumption that the word has ended.

Parameters:

	wordLength	the length of each classified word is saved here.
	wordEndScores	the score of the word end hypotheses is saved here.

Returns:

vector describing a sequence of word ids.

ivector lti::hmmOnlineClassifier::fullTraceBack

(

)

const

Full word sequence traceback Return the best word sequence, given the assumption that the word has ended.

Returns:

vector describing a sequence of word ids.

int lti::hmmOnlineClassifier::getNumberOfHypotheses

(

)

const [inline]

Returns:

number of currently active hypotheses.

const parameters& lti::hmmOnlineClassifier::getParameters

(

)

const

returns used parameters

Reimplemented from lti::hmmClassifier.

virtual const char* lti::hmmOnlineClassifier::getTypeName

(

)

const [virtual]

returns the name of this type ("hmmOnlineClassifier")

Reimplemented from lti::hmmClassifier.

void lti::hmmOnlineClassifier::initialize

(

)

Empty traceback fields and delete current hypotheses.

This routine must be called before a new online classification session is started.

Reimplemented from lti::object.

hmmOnlineClassifier& lti::hmmOnlineClassifier::operator=

(

const hmmOnlineClassifier &

other

)

alias for copy member

Parameters:

other

the classifier to be copied

Returns:

a reference to this classifier object

Reimplemented from lti::hmmClassifier.

ivector lti::hmmOnlineClassifier::partialTraceBack	(	ivector &	wordLength,
		dvector &	wordEndScores
	)

Perform partial traceback.

Find the latest timeframe left to where the word sequence is fixed, i.e. find the start of a sentence where later hypotheses have no effect on. This word sequence is returned and the internal traceback fields are truncated accordingly. Furthermore, the internal scores of all current hypotheses are reduced.

Parameters:

	wordLength	the length of each classified word is saved here.
	wordEndScores	the score of the word end hypotheses is saved here.

Returns:

vector describing a sequence of word ids.

ivector lti::hmmOnlineClassifier::partialTraceBack

(

)

Perform partial traceback.

Find the latest timeframe left to where the word sequence is fixed, i.e. find the start of a sentence where later hypotheses have no effect on. This word sequence is returned and the internal traceback fields are truncated accordingly. Furthermore, the internal scores of all current hypotheses are reduced.

Returns:

vector describing a sequence of word ids.

virtual bool lti::hmmOnlineClassifier::read	(	ioHandler &	handler,
		const bool	complete = true
	)			[virtual]

read method

Reimplemented from lti::hmmClassifier.

void lti::hmmOnlineClassifier::reset

(

)

delete all models, hypotheses and traceback fields.

Reimplemented from lti::hmmClassifier.

int lti::hmmOnlineClassifier::wordErrorCount	(	const ivector &	wordSequence1,
		const ivector &	wordSequence2
	)			const

Calculate the Levenshtein distance (also called edit distance).

The Levenshtein distance of two sentences is the minimum number of substitutions, insertions and deletions needed to transform one sentence into the other.

Returns:

Levenshtein distance of the given word sequences

Referenced by wordErrorRate().

float lti::hmmOnlineClassifier::wordErrorRate	(	const ivector &	correctSequence,
		const ivector &	classifiedSequence
	)			const [inline]

The word error rate is defined as the Levenshtein distance divided by the length of the correct word sequence.

Returns:

word error rate

References lti::genericVector< T >::size(), and wordErrorCount().

virtual bool lti::hmmOnlineClassifier::write	(	ioHandler &	handler,
		const bool	complete = true
	)			const [virtual]

write method

Reimplemented from lti::hmmClassifier.

---

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

• ltiHmmOnlineClassifier.h