org.deeplearning4j.models.embeddings.reader.impl

Class BasicModelUtils<T extends SequenceElement>

java.lang.Object

org.deeplearning4j.models.embeddings.reader.impl.BasicModelUtils<T>

All Implemented Interfaces:

ModelUtils<T>

Direct Known Subclasses:

FlatModelUtils

public class BasicModelUtils<T extends SequenceElement>
extends Object
implements ModelUtils<T>

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	BasicModelUtils.ArrayComparator
static class	BasicModelUtils.SimilarityComparator
static class	BasicModelUtils.WordSimilarity

Field Summary

Fields

Modifier and Type	Field and Description
static String	CORRECT
static String	EXISTS
protected WeightLookupTable<T>	lookupTable
protected boolean	normalized
protected VocabCache<T>	vocabCache
static String	WRONG

Constructor Summary

Constructors

Constructor and Description
BasicModelUtils()

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
protected org.nd4j.linalg.api.ndarray.INDArray	adjustRank(org.nd4j.linalg.api.ndarray.INDArray words)
static List<String>	getLabels(List<BasicModelUtils.WordSimilarity> results, int limit)
void	init(@NonNull WeightLookupTable<T> lookupTable) This method implementations should accept given lookup table, and use them in further calls to interface methods
double	similarity(@NonNull String label1, @NonNull 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
Collection<String>	wordsNearest(@NonNull Collection<String> positive, @NonNull Collection<String> negative, int top) Words nearest based on positive and negative words
Collection<String>	wordsNearest(org.nd4j.linalg.api.ndarray.INDArray words, int top) Words nearest based on positive and negative words * @param top the top n words
Collection<String>	wordsNearest(String label, int n) This method implementations should return N nearest elements labels to given element's label
Collection<String>	wordsNearestSum(Collection<String> positive, Collection<String> negative, int top) Words nearest based on positive and negative words
Collection<String>	wordsNearestSum(org.nd4j.linalg.api.ndarray.INDArray words, int top) Words nearest based on positive and negative words * @param top the top n words
Collection<String>	wordsNearestSum(String word, int n) Get the top n words most similar to the given word

Methods inherited from class java.lang.Object

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

Field Detail

EXISTS

public static final String EXISTS

See Also:

Constant Field Values

CORRECT

public static final String CORRECT

See Also:

Constant Field Values

WRONG

public static final String WRONG

See Also:

Constant Field Values

vocabCache

protected volatile VocabCache<T extends SequenceElement> vocabCache

lookupTable

protected volatile WeightLookupTable<T extends SequenceElement> lookupTable

normalized

protected volatile boolean normalized

Constructor Detail

BasicModelUtils

public BasicModelUtils()

Method Detail

init

public void init(@NonNull
                 @NonNull WeightLookupTable<T> lookupTable)

Description copied from interface: ModelUtils

This method implementations should accept given lookup table, and use them in further calls to interface methods

Specified by:

init in interface ModelUtils<T extends SequenceElement>

similarity

public double similarity(@NonNull
                         @NonNull String label1,
                         @NonNull
                         @NonNull String label2)

Returns the similarity of 2 words. Result value will be in range [-1,1], where -1.0 is exact opposite similarity, i.e. NO similarity, and 1.0 is total match of two word vectors. However, most of time you'll see values in range [0,1], but that's something depends of training corpus. Returns NaN if any of labels not exists in vocab, or any label is null

Specified by:

similarity in interface ModelUtils<T extends SequenceElement>

Parameters:

label1 - the first word

label2 - the second word

Returns:

a normalized similarity (cosine similarity)

wordsNearest

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

Description copied from interface: ModelUtils

This method implementations should return N nearest elements labels to given element's label

Specified by:

wordsNearest in interface ModelUtils<T extends SequenceElement>

Parameters:

label - label to return nearest elements for

n - number of nearest words to return

Returns:

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

Specified by:

accuracy in interface ModelUtils<T extends SequenceElement>

Parameters:

questions - the questions to ask

Returns:

the accuracy based on these questions

similarWordsInVocabTo

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

Find all words with a similar characters in the vocab

Specified by:

similarWordsInVocabTo in interface ModelUtils<T extends SequenceElement>

Parameters:

word - the word to compare

accuracy - the accuracy: 0 to 1

Returns:

the list of words that are similar in the vocab

wordsNearest

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

Description copied from interface: ModelUtils

Words nearest based on positive and negative words

Specified by:

wordsNearest in interface ModelUtils<T extends SequenceElement>

Parameters:

positive - the positive words

negative - the negative words

top - the top n words

Returns:

the words nearest the mean of the words

wordsNearestSum

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

Get the top n words most similar to the given word

Specified by:

wordsNearestSum in interface ModelUtils<T extends SequenceElement>

Parameters:

word - the word to compare

n - the n to get

Returns:

the top n words

adjustRank

protected org.nd4j.linalg.api.ndarray.INDArray adjustRank(org.nd4j.linalg.api.ndarray.INDArray words)

wordsNearest

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

Words nearest based on positive and negative words * @param top the top n words

Specified by:

wordsNearest in interface ModelUtils<T extends SequenceElement>

Returns:

the words nearest the mean of the words

wordsNearestSum

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

Words nearest based on positive and negative words * @param top the top n words

Specified by:

wordsNearestSum in interface ModelUtils<T extends SequenceElement>

Returns:

the words nearest the mean of the words

wordsNearestSum

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

Words nearest based on positive and negative words

Specified by:

wordsNearestSum in interface ModelUtils<T extends SequenceElement>

Parameters:

positive - the positive words

negative - the negative words

top - the top n words

Returns:

the words nearest the mean of the words

getLabels

public static List<String> getLabels(List<BasicModelUtils.WordSimilarity> results,
                                     int limit)