edu.northwestern.at.morphadorner.corpuslinguistics.postagger.smoothing.contextual

Class AbstractContextualSmoother

java.lang.Object

edu.northwestern.at.utils.IsCloseableObject

edu.northwestern.at.morphadorner.corpuslinguistics.postagger.smoothing.contextual.AbstractContextualSmoother

All Implemented Interfaces:

ContextualSmoother, UsesLogger

Direct Known Subclasses:

AdditiveContextualSmoother, DeletedInterpolationContextualSmoother

public abstract class AbstractContextualSmoother
extends IsCloseableObject
implements ContextualSmoother, UsesLogger

Abstract contextual smoother.

An abstract contextual smoother which provides implementations of common service methods such as setting the lexicon and the transition probability matrix. Extend this class and override the abstract method "contextualProbability" to produce a new contextual smoother.

Field Summary

Fields

Modifier and Type	Field and Description
protected Map3D<java.lang.String,java.lang.String,java.lang.String,Probability>	cachedContextualProbabilities Cached contextual probabilities.
protected Logger	logger Logger used for output.
protected PartOfSpeechTagger	partOfSpeechTagger The part of speech tagger for which this smoother provides amoother contextual probabilities.

Constructor Summary

Constructors

Constructor and Description
AbstractContextualSmoother() Create an abstract contextual smoother.

Method Summary

Methods

Modifier and Type	Method and Description
int	cachedProbabilitiesCount() Get the number of cached contextual probabilities.
void	clearCachedProbabilities() Clear cached probabilities..
abstract Probability	contextualProbability(java.lang.String word, java.lang.String tag) Get contextually smoothed probability of a word given a tag.
Logger	getLogger() Get the logger.
void	setLogger(Logger logger) Set the logger.
void	setPartOfSpeechTagger(PartOfSpeechTagger partOfSpeechTagger) Set the part of speech tagger for this smoother.

Methods inherited from class edu.northwestern.at.utils.IsCloseableObject

close

Methods inherited from class java.lang.Object

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

Methods inherited from interface edu.northwestern.at.morphadorner.corpuslinguistics.postagger.smoothing.contextual.ContextualSmoother

contextualProbability

Field Detail

partOfSpeechTagger

protected PartOfSpeechTagger partOfSpeechTagger

The part of speech tagger for which this smoother provides amoother contextual probabilities.

cachedContextualProbabilities

protected Map3D<java.lang.String,java.lang.String,java.lang.String,Probability> cachedContextualProbabilities

Cached contextual probabilities.

logger

protected Logger logger

Logger used for output.

Constructor Detail

AbstractContextualSmoother

public AbstractContextualSmoother()

Create an abstract contextual smoother.

Method Detail

getLogger

public Logger getLogger()

Get the logger.

Specified by:

getLogger in interface UsesLogger

Returns:

The logger.

setLogger

public void setLogger(Logger logger)

Set the logger.

Specified by:

setLogger in interface UsesLogger

Parameters:

logger - The logger.

setPartOfSpeechTagger

public void setPartOfSpeechTagger(PartOfSpeechTagger partOfSpeechTagger)

Set the part of speech tagger for this smoother.

Specified by:

setPartOfSpeechTagger in interface ContextualSmoother

Parameters:

partOfSpeechTagger - Part of speech tagger for which this smoother provides probabilities.

cachedProbabilitiesCount

public int cachedProbabilitiesCount()

Get the number of cached contextual probabilities.

Specified by:

cachedProbabilitiesCount in interface ContextualSmoother

Returns:

The number of cached contextual probabilities.

clearCachedProbabilities

public void clearCachedProbabilities()

Clear cached probabilities..

Specified by:

clearCachedProbabilities in interface ContextualSmoother

contextualProbability

public abstract Probability contextualProbability(java.lang.String word,
                                java.lang.String tag)

Get contextually smoothed probability of a word given a tag.

Specified by:

contextualProbability in interface ContextualSmoother

Parameters:

word - The word.

tag - The part of speech tag.

Returns:

Contextually smoothed probability of word given tag, e.g., p( word | tag ).

To avoid redoing potentially expensive probability calculations, you can use the "cachedContextualProbabilities" HashMap2D to store probabilities once they are calculated. Your contextualProbability method should look to see if the cache contains the needed contextual probability. If so, just retrieve it without recomputing it. If the cache does not contain the probability, compute it, and store it in the cache for future use.

Here is what a contextualProbability method should look like, using the cache.

  protected Probability contextualProbability( String word , String tag )
  {
                              //  See if the contextual probability
                              //  p( word | tag ) is in the cache.

      Probability result  =
          (Probability)cachedContextualProbabilities.get( word , tag );

                              //  If the probability isn't in the
                              //  cache, compute it.

      if ( result == null )
      {
          double prob     = compute smoothed probability value

                              //  Store computed probability in
                              //  the cache.

          result  = new Probability( prob );

          cachedContextualProbabilities.put( word , tag , result );
      }

      return result;
  }