it.uniroma2.sag.kelp.wordspace

Class Wordspace

java.lang.Object

it.uniroma2.sag.kelp.wordspace.Wordspace

All Implemented Interfaces:

WordspaceI

public class Wordspace
extends Object
implements WordspaceI

This is an implementation of a wordspace used for associating words to vectors. In particular this wordspace represents each word as a DenseVector

NOTE: in order to speed-up the computation and to reduce the memory occupation, vectors, instead of being associated to words, are associated to their MD5. This, in some remote cases, can lead to word-collision. If it happens when the wordspace is loaded a WARNING message is provided.

Author:

Danilo Croce, Simone Filice

Constructor Summary

Constructors

Constructor and Description
Wordspace()
Wordspace(String matrixPath)

Method Summary

Methods

Modifier and Type	Method and Description
void	addWordVector(String word, Vector vector) Stores the vector associated to a word.
char[][]	getDictionaryDanilo() Returns the complete set of words in the vocabulary (words having an associated vector in this wordspace)
String	getMatrixPath()
Vector	getVector(String word) Returns the vector associated to the given word
Vector	getZeroVector() Returns the zero vector in the wordspace, i.e, a zero vector having the worspace dimensionality
void	setMatrixPath(String matrixPath) Sets the path of the file where the word vectors are stored and loads them.

Methods inherited from class java.lang.Object

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

Constructor Detail

Wordspace

public Wordspace()

Wordspace

public Wordspace(String matrixPath)
          throws IOException

Throws:

IOException

Method Detail

addWordVector

public void addWordVector(String word,
                 Vector vector)

Description copied from interface: WordspaceI

Stores the vector associated to a word. This vector will be returned from the method getVector

Specified by:

addWordVector in interface WordspaceI

Parameters:

word - the word

vector - the vector associated to word

getVector

public Vector getVector(String word)

Description copied from interface: WordspaceI

Returns the vector associated to the given word

Specified by:

getVector in interface WordspaceI

Parameters:

word - the word whose corresponding vector must be retrieved

Returns:

the vector associated to word, null if word is not in the vocabulary of this wordspace

getDictionaryDanilo

public char[][] getDictionaryDanilo()

Description copied from interface: WordspaceI

Returns the complete set of words in the vocabulary (words having an associated vector in this wordspace)

Specified by:

getDictionaryDanilo in interface WordspaceI

Returns:

the set of all the words in the vocabulary

getMatrixPath

public String getMatrixPath()

Returns:

the matrixPath

setMatrixPath

public void setMatrixPath(String matrixPath)
                   throws IOException

Sets the path of the file where the word vectors are stored and loads them. The file can be a plain text file or a .gz archive.

The expected format is:
number_of_vectors space_dimensionality
word_i [TAB] 1.0 [TAB] 0 [TAB] vector values comma separated

Example:

3 5
dog::n [TAB] 1.0 [TAB] 0 [TAB] 2.1,4.1,1.4,2.3,0.9
cat::n [TAB] 1.0 [TAB] 0 [TAB] 3.2,4.3,1.2,2.2,0.8
mouse::n [TAB] 1.0 [TAB] 0 [TAB] 2.4,4.4,2.4,1.3,0.92


Parameters:

matrixPath - the matrixPath to set

Throws:

IOException

getZeroVector

public Vector getZeroVector()

Description copied from interface: WordspaceI

Returns the zero vector in the wordspace, i.e, a zero vector having the worspace dimensionality

Specified by:

getZeroVector in interface WordspaceI

Returns:

the zero vector associated to this wordspace