edu.cmu.sphinx.linguist.flat

Class FlatLinguist

java.lang.Object

edu.cmu.sphinx.linguist.flat.FlatLinguist

All Implemented Interfaces:

Linguist, Configurable

public class FlatLinguist
extends java.lang.Object
implements Linguist, Configurable

A simple form of the linguist.

The flat linguist takes a Grammar graph (as returned by the underlying, configurable grammar), and generates a search graph for this grammar.

It makes the following simplifying assumptions:

• Zero or one word per grammar node
• No fan-in allowed ever
• No composites (yet)
• Only Unit, HMMState, and pronunciation states (and the initial/final grammar state are in the graph (no word, alternative or grammar states attached).
• Only valid transitions (matching contexts) are allowed
• No tree organization of units
• Branching grammar states are allowed

Note that all probabilities are maintained in the log math domain

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected class	FlatLinguist.FlatSearchGraph The search graph that is produced by the flat linguist.
protected class	FlatLinguist.GState This is a nested class that is used to manage the construction of the states in a grammar node.

Field Summary

Fields

Modifier and Type	Field and Description
protected StatisticsVariable	actualArcs
protected boolean	addOutOfGrammarBranch
protected Cache<SentenceHMMStateArc>	arcPool
protected Grammar	grammar
protected GrammarNode	initialGrammarState
protected LogMath	logMath
protected static float	logOne
protected float	logOutOfGrammarBranchProbability
protected float	logPhoneInsertionProbability
protected java.util.Map<GrammarNode,FlatLinguist.GState>	nodeStateMap
protected AcousticModel	phoneLoopAcousticModel
static java.lang.String	PROP_ACOUSTIC_MODEL The property used to define the acoustic model to use when building the search graph
static java.lang.String	PROP_ADD_OUT_OF_GRAMMAR_BRANCH The property that specifies whether to add a branch for detecting out-of-grammar utterances.
static java.lang.String	PROP_DUMP_GSTATES The property used to determine whether or not the gstates are dumped.
static java.lang.String	PROP_GRAMMAR The property used to define the grammar to use when building the search graph
static java.lang.String	PROP_OUT_OF_GRAMMAR_PROBABILITY The property for the probability of entering the out-of-grammar branch.
static java.lang.String	PROP_PHONE_INSERTION_PROBABILITY The property for the probability of inserting a CI phone in the out-of-grammar ci phone loop
static java.lang.String	PROP_PHONE_LOOP_ACOUSTIC_MODEL The property for the acoustic model used for the CI phone loop.
static java.lang.String	PROP_SHOW_COMPILATION_PROGRESS Property to control whether compilation progress is displayed on standard output.
static java.lang.String	PROP_SPREAD_WORD_PROBABILITIES_ACROSS_PRONUNCIATIONS Property that controls whether word probabilities are spread across all pronunciations.
static java.lang.String	PROP_UNIT_MANAGER The property used to define the unit manager to use when building the search graph
protected SearchGraph	searchGraph
protected StatisticsVariable	totalArcs
protected StatisticsVariable	totalStates

Fields inherited from interface edu.cmu.sphinx.linguist.Linguist

PROP_FILLER_INSERTION_PROBABILITY, PROP_LANGUAGE_WEIGHT, PROP_SILENCE_INSERTION_PROBABILITY, PROP_UNIT_INSERTION_PROBABILITY, PROP_WORD_INSERTION_PROBABILITY

Constructor Summary

Constructors

Constructor and Description
FlatLinguist()
FlatLinguist(AcousticModel acousticModel, Grammar grammar, UnitManager unitManager, double wordInsertionProbability, double silenceInsertionProbability, double fillerInsertionProbability, double unitInsertionProbability, float languageWeight, boolean dumpGStates, boolean showCompilationProgress, boolean spreadWordProbabilitiesAcrossPronunciations, boolean addOutOfGrammarBranch, double outOfGrammarBranchProbability, double phoneInsertionProbability, AcousticModel phoneLoopAcousticModel)

Method Summary

Modifier and Type	Method and Description
protected void	addStartingPath() Ensures that there is a starting path by adding an empty left context to the starting gstate
protected void	addStartingPath(GrammarNode initialNode) Start the search at the indicated node
void	allocate() Allocates the linguist.
protected void	allocateAcousticModel() Allocates the acoustic model.
protected java.util.Collection<SentenceHMMState>	compileGrammar() Compiles the grammar into a sentence HMM.
protected FlatLinguist.GState	createGState(GrammarNode grammarNode) Returns a new GState for the given GrammarNode.
void	deallocate() Deallocates the linguist.
protected SentenceHMMState	findStartingState() Finds the starting state
protected SentenceHMMStateArc	getArc(SentenceHMMState nextState, float logLanguageProbability, float logInsertionProbability) Gets a SentenceHMMStateArc.
protected FlatLinguist.GState	getGState(GrammarNode node) Given a grammar node, retrieve the grammar state
float	getLogSilenceInsertionProbability() Returns the log silence insertion probability.
java.lang.String	getName()
SearchGraph	getSearchGraph() Returns the search graph
protected boolean	grammarHasChanged() Determines if the underlying grammar has changed since we last compiled the search graph
void	newProperties(PropertySheet ps) This method is called when this configurable component needs to be reconfigured.
void	startRecognition() Called before a recognition
void	stopRecognition() Called after a recognition

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

PROP_GRAMMAR

@S4Component(type=Grammar.class)
public static final java.lang.String PROP_GRAMMAR

The property used to define the grammar to use when building the search graph

See Also:

Constant Field Values

PROP_UNIT_MANAGER

@S4Component(type=UnitManager.class)
public static final java.lang.String PROP_UNIT_MANAGER

The property used to define the unit manager to use when building the search graph

See Also:

Constant Field Values

PROP_ACOUSTIC_MODEL

@S4Component(type=AcousticModel.class)
public static final java.lang.String PROP_ACOUSTIC_MODEL

The property used to define the acoustic model to use when building the search graph

See Also:

Constant Field Values

PROP_DUMP_GSTATES

@S4Boolean(defaultValue=false)
public static final java.lang.String PROP_DUMP_GSTATES

The property used to determine whether or not the gstates are dumped.

See Also:

Constant Field Values

PROP_ADD_OUT_OF_GRAMMAR_BRANCH

@S4Boolean(defaultValue=false)
public static final java.lang.String PROP_ADD_OUT_OF_GRAMMAR_BRANCH

The property that specifies whether to add a branch for detecting out-of-grammar utterances.

See Also:

Constant Field Values

PROP_OUT_OF_GRAMMAR_PROBABILITY

@S4Double(defaultValue=1.0)
public static final java.lang.String PROP_OUT_OF_GRAMMAR_PROBABILITY

The property for the probability of entering the out-of-grammar branch.

See Also:

Constant Field Values

PROP_PHONE_LOOP_ACOUSTIC_MODEL

@S4Component(type=AcousticModel.class)
public static final java.lang.String PROP_PHONE_LOOP_ACOUSTIC_MODEL

The property for the acoustic model used for the CI phone loop.

See Also:

Constant Field Values

PROP_PHONE_INSERTION_PROBABILITY

@S4Double(defaultValue=1.0)
public static final java.lang.String PROP_PHONE_INSERTION_PROBABILITY

The property for the probability of inserting a CI phone in the out-of-grammar ci phone loop

See Also:

Constant Field Values

PROP_SHOW_COMPILATION_PROGRESS

@S4Boolean(defaultValue=false)
public static final java.lang.String PROP_SHOW_COMPILATION_PROGRESS

Property to control whether compilation progress is displayed on standard output. If this property is true, a 'dot' is displayed for every 1000 search states added to the search space

See Also:

Constant Field Values

PROP_SPREAD_WORD_PROBABILITIES_ACROSS_PRONUNCIATIONS

@S4Boolean(defaultValue=false)
public static final java.lang.String PROP_SPREAD_WORD_PROBABILITIES_ACROSS_PRONUNCIATIONS

Property that controls whether word probabilities are spread across all pronunciations.

See Also:

Constant Field Values

logOne

protected static final float logOne

See Also:

Constant Field Values

grammar

protected Grammar grammar

logMath

protected LogMath logMath

phoneLoopAcousticModel

protected AcousticModel phoneLoopAcousticModel

addOutOfGrammarBranch

protected boolean addOutOfGrammarBranch

logOutOfGrammarBranchProbability

protected float logOutOfGrammarBranchProbability

logPhoneInsertionProbability

protected float logPhoneInsertionProbability

totalStates

protected StatisticsVariable totalStates

totalArcs

protected StatisticsVariable totalArcs

actualArcs

protected StatisticsVariable actualArcs

nodeStateMap

protected java.util.Map<GrammarNode,FlatLinguist.GState> nodeStateMap

arcPool

protected Cache<SentenceHMMStateArc> arcPool

initialGrammarState

protected GrammarNode initialGrammarState

searchGraph

protected SearchGraph searchGraph

Constructor Detail

FlatLinguist

public FlatLinguist(AcousticModel acousticModel,
                    Grammar grammar,
                    UnitManager unitManager,
                    double wordInsertionProbability,
                    double silenceInsertionProbability,
                    double fillerInsertionProbability,
                    double unitInsertionProbability,
                    float languageWeight,
                    boolean dumpGStates,
                    boolean showCompilationProgress,
                    boolean spreadWordProbabilitiesAcrossPronunciations,
                    boolean addOutOfGrammarBranch,
                    double outOfGrammarBranchProbability,
                    double phoneInsertionProbability,
                    AcousticModel phoneLoopAcousticModel)

FlatLinguist

public FlatLinguist()

Method Detail

getSearchGraph

public SearchGraph getSearchGraph()

Returns the search graph

Specified by:

getSearchGraph in interface Linguist

Returns:

the search graph

newProperties

public void newProperties(PropertySheet ps)
                   throws PropertyException

Description copied from interface: Configurable

This method is called when this configurable component needs to be reconfigured.

Specified by:

newProperties in interface Configurable

Parameters:

ps - a property sheet holding the new data

Throws:

PropertyException - if there is a problem with the properties.

getName

public java.lang.String getName()

allocate

public void allocate()
              throws java.io.IOException

Description copied from interface: Linguist

Allocates the linguist. Resources allocated by the linguist are allocated here. This method may take many seconds to complete depending upon the linguist.

Implementor's Note - A well written linguist will allow allocate to be called multiple times without harm. This will allow a linguist to be shared by multiple search managers.

Specified by:

allocate in interface Linguist

Throws:

java.io.IOException - if an IO error occurs

allocateAcousticModel

protected void allocateAcousticModel()
                              throws java.io.IOException

Allocates the acoustic model.

Throws:

java.io.IOException - if error occured

deallocate

public void deallocate()

Description copied from interface: Linguist

Deallocates the linguist. Any resources allocated by this linguist are released.

Implementor's Note - if the linguist is being shared by multiple searches, the deallocate should only actually deallocate things when the last call to deallocate is made. Two approaches for dealing with this:

(1) Keep an allocation counter that is incremented during allocate and decremented during deallocate. Only when the counter reaches zero should the actually deallocation be performed.

(2) Do nothing in dellocate - just the the GC take care of things

Specified by:

deallocate in interface Linguist

startRecognition

public void startRecognition()

Called before a recognition

Specified by:

startRecognition in interface Linguist

stopRecognition

public void stopRecognition()

Called after a recognition

Specified by:

stopRecognition in interface Linguist

getLogSilenceInsertionProbability

public float getLogSilenceInsertionProbability()

Returns the log silence insertion probability.

Returns:

the log silence insertion probability.

compileGrammar

protected java.util.Collection<SentenceHMMState> compileGrammar()

Compiles the grammar into a sentence HMM. A GrammarJob is created for the initial grammar node and added to the GrammarJob queue. While there are jobs left on the grammar job queue, a job is removed from the queue and the associated grammar node is expanded and attached to the tails. GrammarJobs for the successors are added to the grammar job queue.

Returns:

collection of states

createGState

protected FlatLinguist.GState createGState(GrammarNode grammarNode)

Returns a new GState for the given GrammarNode.

Parameters:

grammarNode - grammar node

Returns:

a new GState for the given GrammarNode

addStartingPath

protected void addStartingPath()

Ensures that there is a starting path by adding an empty left context to the starting gstate

addStartingPath

protected void addStartingPath(GrammarNode initialNode)

Start the search at the indicated node

Parameters:

initialNode - initial node

grammarHasChanged

protected boolean grammarHasChanged()

Determines if the underlying grammar has changed since we last compiled the search graph

Returns:

true if the grammar has changed

findStartingState

protected SentenceHMMState findStartingState()

Finds the starting state

Returns:

the starting state

getArc

protected SentenceHMMStateArc getArc(SentenceHMMState nextState,
                                     float logLanguageProbability,
                                     float logInsertionProbability)

Gets a SentenceHMMStateArc. The arc is drawn from a pool of arcs.

Parameters:

nextState - the next state

logLanguageProbability - the log language probability

logInsertionProbability - the log insertion probability

Returns:

sentence arc

getGState

protected FlatLinguist.GState getGState(GrammarNode node)

Given a grammar node, retrieve the grammar state

Parameters:

node - the grammar node

Returns:

the grammar state associated with the node