org.deeplearning4j.models.word2vec

Class StaticWord2Vec

java.lang.Object

org.deeplearning4j.models.word2vec.StaticWord2Vec

All Implemented Interfaces:

Serializable, WordVectors, org.deeplearning4j.nn.weights.embeddings.EmbeddingInitializer

public class StaticWord2Vec
extends Object
implements WordVectors

This is special limited Word2Vec implementation, suited for serving as lookup table in concurrent multi-gpu environment This implementation DOES NOT load all vectors onto any of gpus, instead of that it holds vectors in, optionally, compressed state in host memory. This implementation DOES NOT provide some of original Word2Vec methods, such as wordsNearest or wordsNearestSum.

Author:

[email protected]

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	StaticWord2Vec.Builder

Method Summary

Modifier and Type	Method and Description
Map<String,Double>	accuracy(List<String> questions) Accuracy based on questions which are a space separated list of strings where the first word is the query word, the next 2 words are negative, and the last word is the predicted word to be nearest PLEASE NOTE: This method is not available in this implementation.
String	getUNK()
double[]	getWordVector(String word) Get the word vector for a given matrix
org.nd4j.linalg.api.ndarray.INDArray	getWordVectorMatrix(String word) Get the word vector for a given matrix
org.nd4j.linalg.api.ndarray.INDArray	getWordVectorMatrixNormalized(String word) Returns the word vector divided by the norm2 of the array
org.nd4j.linalg.api.ndarray.INDArray	getWordVectors(Collection<String> labels) This method returns 2D array, where each row represents corresponding word/label
org.nd4j.linalg.api.ndarray.INDArray	getWordVectorsMean(Collection<String> labels) This method returns mean vector, built from words/labels passed in
boolean	hasWord(String word) Returns true if the model has this word in the vocab
int	indexOf(String word)
protected void	init() Init method validates configuration defined using
boolean	jsonSerializable()
void	loadWeightsInto(org.nd4j.linalg.api.ndarray.INDArray array)
WeightLookupTable	lookupTable() Lookup table for the vectors PLEASE NOTE: This method is not available in this implementation.
boolean	outOfVocabularySupported() Does implementation vectorize words absent in vocabulary
void	setModelUtils(ModelUtils utils) Specifies ModelUtils to be used to access model PLEASE NOTE: This method has no effect in this implementation.
void	setUNK(String newUNK)
double	similarity(String label1, String label2) Returns the similarity of 2 words
List<String>	similarWordsInVocabTo(String word, double accuracy) Find all words with a similar characters in the vocab PLEASE NOTE: This method is not available in this implementation.
int	vectorSize()
VocabCache	vocab() Vocab for the vectors
long	vocabSize()
Collection<String>	wordsNearest(Collection<String> positive, Collection<String> negative, int top) Words nearest based on positive and negative words PLEASE NOTE: This method is not available in this implementation.
Collection<String>	wordsNearest(org.nd4j.linalg.api.ndarray.INDArray words, int top)
Collection<String>	wordsNearest(String word, int n) Get the top n words most similar to the given word PLEASE NOTE: This method is not available in this implementation.
Collection<String>	wordsNearestSum(Collection<String> positive, Collection<String> negative, int top) Words nearest based on positive and negative words PLEASE NOTE: This method is not available in this implementation.
Collection<String>	wordsNearestSum(org.nd4j.linalg.api.ndarray.INDArray words, int top)
Collection<String>	wordsNearestSum(String word, int n) Get the top n words most similar to the given word PLEASE NOTE: This method is not available in this implementation.

Methods inherited from class java.lang.Object

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

Method Detail

getUNK

public String getUNK()

Specified by:

getUNK in interface WordVectors

setUNK

public void setUNK(String newUNK)

Specified by:

setUNK in interface WordVectors

init

protected void init()

Init method validates configuration defined using

hasWord

public boolean hasWord(String word)

Returns true if the model has this word in the vocab

Specified by:

hasWord in interface WordVectors

Parameters:

word - the word to test for

Returns:

true if the model has the word in the vocab

wordsNearest

public Collection<String> wordsNearest(org.nd4j.linalg.api.ndarray.INDArray words,
                                       int top)

Specified by:

wordsNearest in interface WordVectors

wordsNearestSum

public Collection<String> wordsNearestSum(org.nd4j.linalg.api.ndarray.INDArray words,
                                          int top)

Specified by:

wordsNearestSum in interface WordVectors

wordsNearestSum

public Collection<String> wordsNearestSum(String word,
                                          int n)

Get the top n words most similar to the given word PLEASE NOTE: This method is not available in this implementation.

Specified by:

wordsNearestSum in interface WordVectors

Parameters:

word - the word to compare

n - the n to get

Returns:

the top n words

wordsNearestSum

public Collection<String> wordsNearestSum(Collection<String> positive,
                                          Collection<String> negative,
                                          int top)

Words nearest based on positive and negative words PLEASE NOTE: This method is not available in this implementation.

Specified by:

wordsNearestSum in interface WordVectors

Parameters:

positive - the positive words

negative - the negative words

top - the top n words

Returns:

the words nearest the mean of the words

accuracy

public Map<String,Double> accuracy(List<String> questions)

Accuracy based on questions which are a space separated list of strings where the first word is the query word, the next 2 words are negative, and the last word is the predicted word to be nearest PLEASE NOTE: This method is not available in this implementation.

Specified by:

accuracy in interface WordVectors

Parameters:

questions - the questions to ask

Returns:

the accuracy based on these questions

indexOf

public int indexOf(String word)

Specified by:

indexOf in interface WordVectors

similarWordsInVocabTo

public List<String> similarWordsInVocabTo(String word,
                                          double accuracy)

Find all words with a similar characters in the vocab PLEASE NOTE: This method is not available in this implementation.

Specified by:

similarWordsInVocabTo in interface WordVectors

Parameters:

word - the word to compare

accuracy - the accuracy: 0 to 1

Returns:

the list of words that are similar in the vocab

getWordVector

public double[] getWordVector(String word)

Get the word vector for a given matrix

Specified by:

getWordVector in interface WordVectors

Parameters:

word - the word to get the matrix for

Returns:

the ndarray for this word

getWordVectorMatrixNormalized

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

Returns the word vector divided by the norm2 of the array

Specified by:

getWordVectorMatrixNormalized in interface WordVectors

Parameters:

word - the word to get the matrix for

Returns:

the looked up matrix

getWordVectorMatrix

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

Get the word vector for a given matrix

Specified by:

getWordVectorMatrix in interface WordVectors

Parameters:

word - the word to get the matrix for

Returns:

the ndarray for this word

getWordVectors

public org.nd4j.linalg.api.ndarray.INDArray getWordVectors(Collection<String> labels)

This method returns 2D array, where each row represents corresponding word/label

Specified by:

getWordVectors in interface WordVectors

Parameters:

labels -

Returns:

getWordVectorsMean

public org.nd4j.linalg.api.ndarray.INDArray getWordVectorsMean(Collection<String> labels)

This method returns mean vector, built from words/labels passed in

Specified by:

getWordVectorsMean in interface WordVectors

Parameters:

labels -

Returns:

wordsNearest

public Collection<String> wordsNearest(Collection<String> positive,
                                       Collection<String> negative,
                                       int top)

Words nearest based on positive and negative words PLEASE NOTE: This method is not available in this implementation.

Specified by:

wordsNearest in interface WordVectors

Parameters:

positive - the positive words

negative - the negative words

top - the top n words

Returns:

the words nearest the mean of the words

wordsNearest

public Collection<String> wordsNearest(String word,
                                       int n)

Get the top n words most similar to the given word PLEASE NOTE: This method is not available in this implementation.

Specified by:

wordsNearest in interface WordVectors

Parameters:

word - the word to compare

n - the n to get

Returns:

the top n words

similarity

public double similarity(String label1,
                         String label2)

Returns the similarity of 2 words

Specified by:

similarity in interface WordVectors

Parameters:

label1 - the first word

label2 - the second word

Returns:

a normalized similarity (cosine similarity)

vocab

public VocabCache vocab()

Vocab for the vectors

Specified by:

vocab in interface WordVectors

Returns:

lookupTable

public WeightLookupTable lookupTable()

Lookup table for the vectors PLEASE NOTE: This method is not available in this implementation.

Specified by:

lookupTable in interface WordVectors

Returns:

setModelUtils

public void setModelUtils(ModelUtils utils)

Specifies ModelUtils to be used to access model PLEASE NOTE: This method has no effect in this implementation.

Specified by:

setModelUtils in interface WordVectors

Parameters:

utils -

loadWeightsInto

public void loadWeightsInto(org.nd4j.linalg.api.ndarray.INDArray array)

Specified by:

loadWeightsInto in interface org.deeplearning4j.nn.weights.embeddings.EmbeddingInitializer

vocabSize

public long vocabSize()

Specified by:

vocabSize in interface org.deeplearning4j.nn.weights.embeddings.EmbeddingInitializer

vectorSize

public int vectorSize()

Specified by:

vectorSize in interface org.deeplearning4j.nn.weights.embeddings.EmbeddingInitializer

jsonSerializable

public boolean jsonSerializable()

Specified by:

jsonSerializable in interface org.deeplearning4j.nn.weights.embeddings.EmbeddingInitializer

outOfVocabularySupported

public boolean outOfVocabularySupported()

Description copied from interface: WordVectors

Does implementation vectorize words absent in vocabulary

Specified by:

outOfVocabularySupported in interface WordVectors

Returns:

boolean