org.apache.ctakes.relationextractor.ae.features

Class EmbeddingFeatureExtractor

java.lang.Object

org.apache.ctakes.relationextractor.ae.features.EmbeddingFeatureExtractor

All Implemented Interfaces:

RelationFeaturesExtractor<IdentifiedAnnotation,IdentifiedAnnotation>

public class EmbeddingFeatureExtractor
extends Object
implements RelationFeaturesExtractor<IdentifiedAnnotation,IdentifiedAnnotation>

Word embedding based features. OOV words are handled by the average vector which should be included with the rest of the vectors and indexed as "oov".

Constructor Summary

Constructors

Constructor and Description
EmbeddingFeatureExtractor(Map<String,List<Double>> wordVectors)

Method Summary

Modifier and Type	Method and Description
List<Double>	addVectors(List<Double> vector1, List<Double> vector2) Add two vectors.
double	computeCosineSimilarity(List<Double> vector1, List<Double> vector2) Compute cosine similarity between two vectors.
List<org.cleartk.ml.Feature>	extract(org.apache.uima.jcas.JCas jCas, IdentifiedAnnotation arg1, IdentifiedAnnotation arg2) Extract features for the pair of named entity mentions.

Methods inherited from class java.lang.Object

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

Constructor Detail

EmbeddingFeatureExtractor

public EmbeddingFeatureExtractor(Map<String,List<Double>> wordVectors)

Method Detail

extract

public List<org.cleartk.ml.Feature> extract(org.apache.uima.jcas.JCas jCas,
                                            IdentifiedAnnotation arg1,
                                            IdentifiedAnnotation arg2)
                                     throws org.apache.uima.analysis_engine.AnalysisEngineProcessException

Description copied from interface: RelationFeaturesExtractor

Extract features for the pair of named entity mentions.

Specified by:

extract in interface RelationFeaturesExtractor<IdentifiedAnnotation,IdentifiedAnnotation>

Parameters:

jCas - The JCas containing the two named entity mentions.

arg1 - The first identified annotation in the text.

arg2 - The second identified annotation in the text.

Returns:

A list of features indicative of the relation between the named entities

Throws:

org.apache.uima.analysis_engine.AnalysisEngineProcessException

computeCosineSimilarity

public double computeCosineSimilarity(List<Double> vector1,
                                      List<Double> vector2)

Compute cosine similarity between two vectors.

addVectors

public List<Double> addVectors(List<Double> vector1,
                               List<Double> vector2)

Add two vectors. Return the sum vector.