com.datumbox.framework.applications.nlp

Class TextClassifier

java.lang.Object

com.datumbox.framework.core.machinelearning.common.abstracts.AbstractTrainer<MP,TP>

com.datumbox.framework.core.machinelearning.common.abstracts.modelers.AbstractModeler<Modeler.ModelParameters,Modeler.TrainingParameters>

com.datumbox.framework.applications.datamodeling.Modeler

com.datumbox.framework.applications.nlp.TextClassifier

All Implemented Interfaces:

Savable, Parallelizable, Trainable<Modeler.ModelParameters,Modeler.TrainingParameters>, java.lang.AutoCloseable

public class TextClassifier
extends Modeler

TextClassifier is a convenience class which can be used to train Text ClassificationMetrics models. It is a wrapper class which automatically takes care of the text parsing, tokenization, feature selection and modeler training processes. It takes as input either a Dataframe object or multiple text files (one for each category) with one observation per row.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	TextClassifier.ModelParameters It contains all the Model AbstractParameters which are learned during the training.
static class	TextClassifier.TrainingParameters It contains the Training AbstractParameters of the Text Classifier.

Nested classes/interfaces inherited from class com.datumbox.framework.core.machinelearning.common.abstracts.AbstractTrainer

AbstractTrainer.AbstractModelParameters, AbstractTrainer.AbstractTrainingParameters

Method Summary

Modifier and Type	Method and Description
void	fit(java.util.Map<java.lang.Object,java.net.URI> datasets) Trains a Machine Learning modeler using the provided dataset files.
Record	predict(java.lang.String text) It generates a prediction for a particular string.
Dataframe	predict(java.net.URI datasetURI) Generates a Dataframe with the predictions for the provided data file.
ClassificationMetrics	validate(Dataframe testDataset) It validates the modeler using the provided dataset and it returns the ClassificationMetrics.
ClassificationMetrics	validate(java.util.Map<java.lang.Object,java.net.URI> datasets) It validates the modeler using the provided dataset files.

Methods inherited from class com.datumbox.framework.applications.datamodeling.Modeler

_predict, close, delete, isParallelized, save, setParallelized

Methods inherited from class com.datumbox.framework.core.machinelearning.common.abstracts.modelers.AbstractModeler

predict

Methods inherited from class com.datumbox.framework.core.machinelearning.common.abstracts.AbstractTrainer

fit, getModelParameters, getTrainingParameters

Methods inherited from class java.lang.Object

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

Method Detail

fit

public void fit(java.util.Map<java.lang.Object,java.net.URI> datasets)

Trains a Machine Learning modeler using the provided dataset files. The data map should have as index the names of each class and as values the URIs of the training files. The training files should contain one training example per row.

Parameters:

datasets -

predict

public Dataframe predict(java.net.URI datasetURI)

Generates a Dataframe with the predictions for the provided data file. The data file should contain the text of one observation per row.

Parameters:

datasetURI -

Returns:

predict

public Record predict(java.lang.String text)

It generates a prediction for a particular string. It returns a Record object which contains the observation data, the predicted class and probabilities.

Parameters:

text -

Returns:

validate

public ClassificationMetrics validate(Dataframe testDataset)

It validates the modeler using the provided dataset and it returns the ClassificationMetrics. The testDataset should contain the real target variables.

Parameters:

testDataset -

Returns:

validate

public ClassificationMetrics validate(java.util.Map<java.lang.Object,java.net.URI> datasets)

It validates the modeler using the provided dataset files. The data map should have as index the names of each class and as values the URIs of the training files. The data files should contain one example per row.

Parameters:

datasets -

Returns: