smile.sequence

Class CRFLabeler<T>

java.lang.Object

smile.sequence.CRFLabeler<T>

All Implemented Interfaces:

java.io.Serializable, SequenceLabeler<T>

public class CRFLabeler<T>
extends java.lang.Object
implements SequenceLabeler<T>

First-order CRF sequence labeler.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
java.util.function.Function<T,smile.data.Tuple>	features The feature function.
CRF	model The CRF model.

Constructor Summary

Constructors

Constructor and Description
CRFLabeler(CRF model, java.util.function.Function<T,smile.data.Tuple> features) Constructor.

Method Summary

Modifier and Type	Method and Description
static <T> CRFLabeler<T>	fit(T[][] sequences, int[][] labels, java.util.function.Function<T,smile.data.Tuple> features) Fits a CRF model.
static <T> CRFLabeler<T>	fit(T[][] sequences, int[][] labels, java.util.function.Function<T,smile.data.Tuple> features, int ntrees, int maxDepth, int maxNodes, int nodeSize, double shrinkage) Fits a CRF.
static <T> CRFLabeler<T>	fit(T[][] sequences, int[][] labels, java.util.function.Function<T,smile.data.Tuple> features, java.util.Properties prop) Fits a CRF model.
int[]	predict(T[] o) Returns the most likely label sequence given the feature sequence by the forward-backward algorithm.
java.lang.String	toString()
int[]	viterbi(T[] o) Labels sequence with Viterbi algorithm.

Methods inherited from class java.lang.Object

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

Field Detail

model

public final CRF model

The CRF model.

features

public final java.util.function.Function<T,smile.data.Tuple> features

The feature function.

Constructor Detail

CRFLabeler

public CRFLabeler(CRF model,
                  java.util.function.Function<T,smile.data.Tuple> features)

Constructor.

Parameters:

model - the CRF model.

features - the feature function.

Method Detail

fit

public static <T> CRFLabeler<T> fit(T[][] sequences,
                                    int[][] labels,
                                    java.util.function.Function<T,smile.data.Tuple> features)

Fits a CRF model.

Parameters:

sequences - the training data.

labels - the training sequence labels.

fit

public static <T> CRFLabeler<T> fit(T[][] sequences,
                                    int[][] labels,
                                    java.util.function.Function<T,smile.data.Tuple> features,
                                    java.util.Properties prop)

Fits a CRF model.

Parameters:

sequences - the training data.

labels - the training sequence labels.

fit

public static <T> CRFLabeler<T> fit(T[][] sequences,
                                    int[][] labels,
                                    java.util.function.Function<T,smile.data.Tuple> features,
                                    int ntrees,
                                    int maxDepth,
                                    int maxNodes,
                                    int nodeSize,
                                    double shrinkage)

Fits a CRF.

Parameters:

sequences - the observation sequences.

labels - the state labels of observations, of which states take values in [0, k), where k is the number of hidden states.

features - the feature function.

ntrees - the number of trees/iterations.

maxDepth - the maximum depth of the tree.

maxNodes - the maximum number of leaf nodes in the tree.

nodeSize - the number of instances in a node below which the tree will not split, setting nodeSize = 5 generally gives good results.

shrinkage - the shrinkage parameter in (0, 1] controls the learning rate of procedure.

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object

predict

public int[] predict(T[] o)

Returns the most likely label sequence given the feature sequence by the forward-backward algorithm.

Specified by:

predict in interface SequenceLabeler<T>

Parameters:

o - an observation sequence.

Returns:

the most likely state sequence.

viterbi

public int[] viterbi(T[] o)

Labels sequence with Viterbi algorithm. Viterbi algorithm returns the whole sequence label that has the maximum probability, which makes sense in applications (e.g.part-of-speech tagging) that require coherent sequential labeling. The forward-backward algorithm labels a sequence by individual prediction on each position. This usually produces better accuracy although the results may not be coherent.