de.julielab.jcore.ae.jnet.tagger

Class NETagger

java.lang.Object

de.julielab.jcore.ae.jnet.tagger.NETagger

public class NETagger
extends Object

general class which does all the ML stuff TODO confidence estimation also for IOB (not only IO)

Author:

tomanek

Constructor Summary

Constructors

Constructor and Description
NETagger() default constructor
NETagger(File featureConfigFile) constructor for feature config file

Method Summary

Modifier and Type	Method and Description
Properties	getFeatureConfig()
Object	getModel() return the model
int	getNumber_Iterations()
boolean	is_Max_Ent()
boolean	isTrained() returns true when model has been successfully trained.
Sentence	PPDtoUnits(String sentence) takes a sentence in piped format and returns the corresponding unit sentence as a Sentence object
void	predict(Sentence sentence, boolean showSegmentConfidence) predicts the entity labels by means of a model.
ArrayList<String>	predictIOB(ArrayList<Sentence> sentences, boolean showSegmentConfidence) predict the entity labels by means of a previously learned model.
void	readModel(File f) load a previously trained FeatureSubsetModel (CRF4+Properties) which was stored as serialized object to disk.
void	readModel(InputStream is) load a previously trained FeatureSubsetModel (CRF4+Properties) which was stored as serialized object to disk.
void	set_Max_Ent(boolean me_train)
void	set_Number_Iterations(int number_iter)
void	setFeatureConfig(Properties featureConfig)
void	train(ArrayList<Sentence> sentences) this is to train a NE model (based on CRF); when trained, the model is stored internally.
void	writeModel(String filename) Save the model learned to disk.

Methods inherited from class java.lang.Object

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

Constructor Detail

NETagger

public NETagger()

default constructor

NETagger

public NETagger(File featureConfigFile)

constructor for feature config file

Parameters:

featureConfigFile -

Method Detail

isTrained

public boolean isTrained()

returns true when model has been successfully trained.

Returns:

true if trained

train

public void train(ArrayList<Sentence> sentences)

this is to train a NE model (based on CRF); when trained, the model is stored internally. The model can be saved to disk using the writeModel command.

Parameters:

sentences - training data, an ArrayList of Sentence objects, File which contains the feature subset to be used in a text format

predict

public void predict(Sentence sentence,
                    boolean showSegmentConfidence)

predicts the entity labels by means of a model. this method is needed by UIMA-JNET!

Parameters:

sentence - a Sentence object containing all units (= tokens) of that sentence

showSegmentConfidence - when this flag is set to true for all found entities a confidence is estimated. The confidence is stored in Unit object.

predictIOB

public ArrayList<String> predictIOB(ArrayList<Sentence> sentences,
                                    boolean showSegmentConfidence)

predict the entity labels by means of a previously learned model. this method is used by JNET stand alone version (for UIMA-JNET see other predict method) Output is an arraylist of IOB

Parameters:

sentences - an ArrayList of Sentence objects

showSegmentConfidence - when this flag is set to true for all found entities a confidence is estimated. Confidence is written to IOB outputfile.

Returns:

IOB output for the sentences to be predicted. Each element of the ArrayList is a string which refers to one word and its label ("token\tlabel")

writeModel

public void writeModel(String filename)

Save the model learned to disk. THis is done via Java's object serialization.

Parameters:

filename - where to write it (full path!)

readModel

public void readModel(File f)
               throws IOException,
                      FileNotFoundException,
                      ClassNotFoundException

load a previously trained FeatureSubsetModel (CRF4+Properties) which was stored as serialized object to disk.

Parameters:

is - input stream of the serialized featureSubsetModel

Throws:

IOException

FileNotFoundException

ClassNotFoundException

readModel

public void readModel(InputStream is)
               throws IOException,
                      FileNotFoundException,
                      ClassNotFoundException

load a previously trained FeatureSubsetModel (CRF4+Properties) which was stored as serialized object to disk.

Parameters:

is - input stream of the serialized featureSubsetModel

Throws:

IOException

FileNotFoundException

ClassNotFoundException

getModel

public Object getModel()

return the model

setFeatureConfig

public void setFeatureConfig(Properties featureConfig)

getFeatureConfig

public Properties getFeatureConfig()

PPDtoUnits

public Sentence PPDtoUnits(String sentence)

takes a sentence in piped format and returns the corresponding unit sentence as a Sentence object

Parameters:

sentence - in piped format to be converted

getNumber_Iterations

public int getNumber_Iterations()

set_Number_Iterations

public void set_Number_Iterations(int number_iter)

is_Max_Ent

public boolean is_Max_Ent()

set_Max_Ent

public void set_Max_Ent(boolean me_train)