de.datexis.sector.tagger

Class SectorEncoder

java.lang.Object

de.datexis.annotator.AnnotatorComponent

de.datexis.encoder.Encoder

de.datexis.sector.tagger.SectorEncoder

All Implemented Interfaces:

IComponent, IEncoder

public class SectorEncoder
extends Encoder

An Encoder that capsules SectorTagger and returns the hidden layer embedding.

Author:

Sebastian Arnold

Field Summary

Fields

Modifier and Type	Field and Description
protected SectorTagger	tagger

Fields inherited from class de.datexis.annotator.AnnotatorComponent

id, log, model, modelAvailable, name, timer

Constructor Summary

Constructors

Constructor and Description
SectorEncoder()
SectorEncoder(String id)
SectorEncoder(String id, SectorTagger sector)

Method Summary

Modifier and Type	Method and Description
org.nd4j.linalg.api.ndarray.INDArray	encode(Span span)
org.nd4j.linalg.api.ndarray.INDArray	encode(String word)
void	encodeEach(Collection<Document> docs, Class<? extends Span> elementClass)
void	encodeEach(Document d, Class<? extends Span> elementClass)
void	encodeEach(Sentence input, Class<? extends Span> elementClass)
org.nd4j.linalg.api.ndarray.INDArray	encodeMatrix(List<Document> input, int maxTimeSteps, Class<? extends Span> timeStepClass) Encodes each element in the input and attaches the vectors to the element.
int	getBatchSize()
int	getEmbeddingLayerSize()
long	getEmbeddingVectorSize()
String	getName()
int	getNumEpochs()
SectorTagger	getTagger()
boolean	isMultiClass()
boolean	isRandomize()
void	loadModel(Resource file) overrides and re-implementations for export of nested tagger to XML/JSON
void	saveModel(Resource dir, String name)
void	setBatchSize(int size)
void	setEmbeddingLayerSize(int size)
void	setMultiClass(boolean isMultiClass)
void	setName(String name)
void	setNumEpochs(int numEpochs)
void	setRandomize(boolean rand)
void	setTagger(SectorTagger tagger)
void	trainModel(Collection<Document> documents)

Methods inherited from class de.datexis.encoder.Encoder

encode, trainModel

Methods inherited from class de.datexis.annotator.AnnotatorComponent

appendTestLog, appendTestLog, appendTrainLog, appendTrainLog, clearTestLog, clearTrainLog, getConf, getId, getModel, getTestLog, getTrainLog, isModelAvailable, isModelAvailableInChildren, setConf, setId, setModel, setModelAvailable, setModelFilename

Methods inherited from class java.lang.Object

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

Methods inherited from interface de.datexis.annotator.IComponent

getEncoders, getId, isModelAvailable, setEncoders

Field Detail

tagger

protected SectorTagger tagger

Constructor Detail

SectorEncoder

public SectorEncoder()

SectorEncoder

public SectorEncoder(String id)

SectorEncoder

public SectorEncoder(String id,
                     SectorTagger sector)

Method Detail

getTagger

public SectorTagger getTagger()

setTagger

public void setTagger(SectorTagger tagger)

getEmbeddingVectorSize

public long getEmbeddingVectorSize()

encode

public org.nd4j.linalg.api.ndarray.INDArray encode(Span span)

encode

public org.nd4j.linalg.api.ndarray.INDArray encode(String word)

encodeEach

public void encodeEach(Document d,
                       Class<? extends Span> elementClass)

Overrides:

encodeEach in class Encoder

encodeEach

public void encodeEach(Collection<Document> docs,
                       Class<? extends Span> elementClass)

Overrides:

encodeEach in class Encoder

encodeMatrix

public org.nd4j.linalg.api.ndarray.INDArray encodeMatrix(List<Document> input,
                                                         int maxTimeSteps,
                                                         Class<? extends Span> timeStepClass)

Encodes each element in the input and attaches the vectors to the element. Please override this if the elements of your encoders are not independent or stateful.

Overrides:

encodeMatrix in class Encoder

Parameters:

input - - the Document that should be encoded

timeStepClass - - the class of sub elements in the Document, e.g. Sentence.class

encodeEach

public void encodeEach(Sentence input,
                       Class<? extends Span> elementClass)

Overrides:

encodeEach in class Encoder

trainModel

public void trainModel(Collection<Document> documents)

Specified by:

trainModel in class Encoder

loadModel

public void loadModel(Resource file)

overrides and re-implementations for export of nested tagger to XML/JSON

saveModel

public void saveModel(Resource dir,
                      String name)

getName

public String getName()

Specified by:

getName in interface IComponent

Overrides:

getName in class AnnotatorComponent

setName

public void setName(String name)

Overrides:

setName in class AnnotatorComponent

getBatchSize

public int getBatchSize()

setBatchSize

public void setBatchSize(int size)

getEmbeddingLayerSize

public int getEmbeddingLayerSize()

setEmbeddingLayerSize

public void setEmbeddingLayerSize(int size)

setMultiClass

public void setMultiClass(boolean isMultiClass)

isMultiClass

public boolean isMultiClass()

setNumEpochs

public void setNumEpochs(int numEpochs)

getNumEpochs

public int getNumEpochs()

setRandomize

public void setRandomize(boolean rand)

isRandomize

public boolean isRandomize()