edu.stanford.nlp.parser.lexparser

Interface UnknownWordModelTrainer

All Known Implementing Classes:

AbstractUnknownWordModelTrainer, ArabicUnknownWordModelTrainer, BaseUnknownWordModelTrainer, ChineseUnknownWordModelTrainer, EnglishUnknownWordModelTrainer, FrenchUnknownWordModelTrainer, GermanUnknownWordModelTrainer, SpanishUnknownWordModelTrainer

public interface UnknownWordModelTrainer

An interface for training an UnknownWordModel. Once initialized, you can feed it trees and then call finishTraining to get the UnknownWordModel.

Author:

John Bauer

Field Summary

Fields

Modifier and Type	Field and Description
static IntTaggedWord	NULL_ITW
static short	nullTag
static int	nullWord
static java.lang.String	unknown

Method Summary

Modifier and Type	Method and Description
UnknownWordModel	finishTraining() Returns the trained UWM.
void	incrementTreesRead(double weight) Maintains a (real-valued) count of how many (weighted) trees have been read in.
void	initializeTraining(Options op, Lexicon lex, Index<java.lang.String> wordIndex, Index<java.lang.String> tagIndex, double totalTrees) Initialize the trainer with a few of the data structures it needs to train.
void	train(java.util.Collection<Tree> trees) Tallies statistics for this particular collection of trees.
void	train(java.util.Collection<Tree> trees, double weight) Tallies statistics for a weighted collection of trees.
void	train(TaggedWord tw, int loc, double weight) Tallies statistics for a single word.
void	train(Tree tree, double weight) Tallies statistics for a single tree.

Field Detail

unknown

static final java.lang.String unknown

See Also:

Constant Field Values

nullWord

static final int nullWord

See Also:

Constant Field Values

nullTag

static final short nullTag

See Also:

Constant Field Values

NULL_ITW

static final IntTaggedWord NULL_ITW

Method Detail

initializeTraining

void initializeTraining(Options op,
                        Lexicon lex,
                        Index<java.lang.String> wordIndex,
                        Index<java.lang.String> tagIndex,
                        double totalTrees)

Initialize the trainer with a few of the data structures it needs to train. Also, it is necessary to estimate the number of trees that it will be given, as many of the UWMs switch training modes after seeing a fraction of the trees. This is an initialization method and not part of the constructor because these Trainers are generally loaded by reflection, and making this a method instead of a constructor lets the compiler catch silly errors.

train

void train(java.util.Collection<Tree> trees)

Tallies statistics for this particular collection of trees. Can be called multiple times.

train

void train(java.util.Collection<Tree> trees,
           double weight)

Tallies statistics for a weighted collection of trees. Can be called multiple times.

train

void train(Tree tree,
           double weight)

Tallies statistics for a single tree. Can be called multiple times.

train

void train(TaggedWord tw,
           int loc,
           double weight)

Tallies statistics for a single word. Can be called multiple times.

incrementTreesRead

void incrementTreesRead(double weight)

Maintains a (real-valued) count of how many (weighted) trees have been read in. Can be called multiple times.

Parameters:

weight - The weight of trees additionally trained on

finishTraining

UnknownWordModel finishTraining()

Returns the trained UWM. Many of the subclasses build exactly one model, and some of the finishTraining methods manipulate the data in permanent ways, so this should only be called once