edu.mit.multimodal.segment
Class AbstractSegmenter

java.lang.Object
  edu.mit.multimodal.segment.AbstractSegmenter

Direct Known Subclasses:

FeatureSegmenter, InterpolatedSegmenter

public abstract class AbstractSegmenter
extends java.lang.Object

Constructor Summary

AbstractSegmenter()

Method Summary

static weka.core.Instances	balance(weka.core.Instances training_instances) Balances a set of boolean-valued instances so that both trues and falses are weighted the same.
void	buildModel(SegmentationTranscript[] p_corpus) Builds a model based on the corpus, including a gesture model and learning the tunable parameters via cross-validation.
java.lang.String	configString() Prints the configuration for this run.
SegmentResultTracker	evaluate() The top-level evaluate call.
static boolean	getBoolFromArgv(java.lang.String[] argv, char flag, boolean def) Uses the weka utility to return a boolean from a command-line flag.
static int	getIntFromArgv(java.lang.String[] argv, char flag, int def) Uses the weka utility to return an integer from a command-line flag.
abstract double[]	getPosteriors(SegmentationTranscript corpus) Returns the combined model posteriors for a given corpus.
static java.lang.String	getStringFromArgv(java.lang.String[] argv, char flag, java.lang.String def) Uses the weka utility to return a String from a command-line flag.
void	initialize(java.lang.String[] argv) Initializes the AbstractSegmenter from argv.
abstract void	makeClassifier(SegmentationTranscript[] p_corpus) Makes the classifier.
static java.lang.String	usageString()

Methods inherited from class java.lang.Object

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

Constructor Detail

AbstractSegmenter

public AbstractSegmenter()

Method Detail

getPosteriors

public abstract double[] getPosteriors(SegmentationTranscript corpus)

Returns the combined model posteriors for a given corpus.

Parameters:

corpus - a SegmentationTranscript value

Returns:

a double[] value

buildModel

public void buildModel(SegmentationTranscript[] p_corpus)

Builds a model based on the corpus, including a gesture model and learning the tunable parameters via cross-validation.

Parameters:

p_corpus - a SegmentationTranscript[] value

makeClassifier

public abstract void makeClassifier(SegmentationTranscript[] p_corpus)

Makes the classifier. Does not learn the tunable parameters.

Parameters:

p_corpus - a SegmentationTranscript[] value

usageString

public static java.lang.String usageString()

evaluate

public SegmentResultTracker evaluate()

The top-level evaluate call. Trains, tunes, and returns results using either cross-validation or randomized test-training splits.

Returns:

a ResultTracker value

balance

public static weka.core.Instances balance(weka.core.Instances training_instances)

Balances a set of boolean-valued instances so that both trues and falses are weighted the same.

Parameters:

training_instances - an Instance value

Returns:

an Instances value

getIntFromArgv

public static int getIntFromArgv(java.lang.String[] argv,
                                 char flag,
                                 int def)

Uses the weka utility to return an integer from a command-line flag.

Parameters:

argv - a String[] value

flag - a char value

def - an int value

Returns:

an int value

getStringFromArgv

public static java.lang.String getStringFromArgv(java.lang.String[] argv,
                                                 char flag,
                                                 java.lang.String def)

Uses the weka utility to return a String from a command-line flag.

Parameters:

argv - a String[] value

flag - a char value

def - a String value

Returns:

a String value

getBoolFromArgv

public static boolean getBoolFromArgv(java.lang.String[] argv,
                                      char flag,
                                      boolean def)

Uses the weka utility to return a boolean from a command-line flag.

Parameters:

argv - a String[] value

flag - a char value

def - a boolean value

Returns:

a boolean value

initialize

public void initialize(java.lang.String[] argv)

Initializes the AbstractSegmenter from argv. Should be called by subclasses.

Parameters:

argv - a String value

configString

public java.lang.String configString()

Prints the configuration for this run.

Returns:

a String value