word2vector

word2vector NodeJS Interface

This is a Node.js interface for Google's word2vector.
Here is an example of how to load large model like GoogleNews-vectors-negative300.bin by this package.

Supports both binary model and raw text model.

Installation

Linux, Unix OS are supported. Install it via npm:

npm install word2vector --save

In Node.js, require the module as below:

var w2v = require( 'word2vector' );

API Document:

---

Overview

train load getVector getVectors getSimilarWords getNeighbors similarity substract add

---

w2v.train( trainFile, modelFile, options, callback )

Click here to see example TrainFile format.
Example:

var w2v = require("./lib");
var trainFile = "./data/train.data",
    modelFile = "./data/test.model.bin";
w2v.train(trainFile, modelFile, {
  	cbow: 1,           // use the continuous bag of words model //default
  	size: 10,          // sets the size (dimension) of word vectors // default 100
  	window: 8,         // sets maximal skip length between words // default 5
    binary: 1,         // save the resulting vectors in binary mode // default off
  	negative: 25,      // number of negative examples; common values are 3 - 10 (0 = not used) // default 5
  	hs: 0,             // 1 = use  Hierarchical Softmax // default 0
  	sample: 1e-4,      
  	threads: 20,
  	iter: 15,
  	minCount: 1,       // This will discard words that appear less than *minCount* times // default 5
    logOn: false       // sets whether any output should be printed to the console // default false
  });

w2v.load( modelFile,?readType = "")

Should load model before calling any calcuation functions.

Params	Description	Default Value
readType	Model format, pass "utf-8" if using a raw text model.	"bin"

var w2v = require("../lib");
var modelFile = "./test.model.bin";
w2v.load( modelFile );
// console.log(w2v.getSimilarWordsWords());

w2v.getVector(word="word")

Params	Description	Default Value
word	String to be searched.	"word"

'use strict';
var w2v = require("./lib");
var modelFile = "./data/test.model.bin";
w2v.load( modelFile );
console.log(w2v.getVector("孫悟空"));
console.log(w2v.getVector("李洵"));

Sample Output:

// Array Type Only
[ 0.104406,
  -0.160019,
  -0.604506,
  -0.622804,
  0.039482,
  -0.120058,
  0.073555,
  0.05646,
  0.099059,
  -0.419282 ]

null // Return null if this word is not in model.

w2v.getVectors(words=["word1", "word2"], ?options = {})

Params	Description	Default Value
words	Array of strings to be searched.	"word"

var w2v = require("./lib");  
var modelFile = "./data/test.model.bin";
w2v.load( modelFile );
console.log(w2v.getVectors(["孫悟空", "李洵"]));

Sample Output:

[ { word: '孫悟空',
    vector:
     [ 0.104406,
       -0.160019,
       -0.604506,
       -0.622804,
       0.039482,
       -0.120058,
       0.073555,
       0.05646,
       0.099059,
       -0.419282 ] },
  { word: '李洵', vector: null } ]
  // this will trigger a error log in console:
  //'李洵' is not found in the model.

w2v.getSimilarWords(word = "word", ?options = {})

Return 40ish words that is similar to "word".

Params	Description	Default Value
word	Strings to be searched.	"word"
options.N	return topN results	Array

var w2v = require("./lib");
var modelFile = "./data/test.model.bin";
w2v.load( modelFile );
console.log(w2v.getSimilarWords("唐三藏"));
console.log(w2v.getSimilarWords("李洵"));

Sample Output:

// Array Type
[ { word: '孫悟空', similarity: 0.974369 },
  { word: '吳承恩', similarity: 0.96686 },
  { word: '林黛玉', similarity: 0.966664 },
  { word: '北地', similarity: 0.96264 },
  { word: '賈寶玉', similarity: 0.962137 },
  { word: '楚霸王', similarity: 0.955795 },
  { word: '梁山泊', similarity: 0.932804 },
  { word: '濮陽', similarity: 0.927542 },
  { word: '黃天霸', similarity: 0.927459 },
  { word: '英雄豪傑', similarity: 0.921575 }]
// Return empty [] if this word is not in model.
'李洵' is not found in the model.
[]

getNeighbors(vector, ?options = {})

Params	Description	Default Value
vector	Vector to be searched.	"word"
options.N	return topN results	Array

var w2v = require("./lib");
var modelFile = "./data/test.model.bin";
w2v.load( modelFile );
var a = w2v.getNeighbors(w2v.getVector("唐三藏"), {N: 9});
// These are equal to use w2v.getSimilarWords("唐三藏");
console.log(a);

Sample Output1:

[ { word: '唐三藏', similarity: 0.9999993515200001 },
  { word: '孫悟空', similarity: 0.974368825898 },
  { word: '吳承恩', similarity: 0.966859435824 },
  { word: '林黛玉', similarity: 0.966663471323 },
  { word: '北地', similarity: 0.962639240211 },
  { word: '賈寶玉', similarity: 0.9621371820049999 },
  { word: '楚霸王', similarity: 0.9557946924850002 },
  { word: '梁山泊', similarity: 0.9328033548890001 },
  { word: '濮陽', similarity: 0.9275417727409999 } ]
{ '唐三藏': 0.9999993515200001,
  '孫悟空': 0.974368825898,
  '吳承恩': 0.966859435824,
  '林黛玉': 0.966663471323,
  '北地': 0.962639240211,
  '賈寶玉': 0.9621371820049999,
  '楚霸王': 0.9557946924850002,
  '梁山泊': 0.9328033548890001,
  '濮陽': 0.9275417727409999 }

w2v.similarity(word1 = "word1", word2 = "word2")

w2v.similarity(vector1 = [], word2 = "word2")

w2v.similarity(word1 = "word1", vector2 = [])

w2v.similarity(vector1 = [], vector2 = [])

Compute the [cosine similarity](https://en.wikipedia.org/wiki/Cosine_similarity) between the two vector.
Will auto search the vector of passed word in model. Return false if it's not found.

Params	Description	Default Value
word1	First Strings to be compared.	No default value
word2	Second Strings to be compared.	No default value
vector1	First Vector to be compared.	No default value
vector2	Second Vector to be compared.	No default value

'use strict';
var w2v = require("./lib");
var modelFile = "./data/test.model.bin";
w2v.load( modelFile );
var a = w2v.similarity("唐三藏", "孫悟空"); //  0.974368825898
console.log(a);
var b = w2v.similarity("唐三藏", "李洵"); //  0.974368825898
// same as var b = w2v.similarity("唐三藏", w2v.getVector("李洵"));
// same as var b = w2v.similarity(w2v.getVector("唐三藏"), "李洵");
// same as var b = w2v.similarity(w2v.getVector("唐三藏"), w2v.getVector("李洵"));
console.log(b);

Sample Output:

0.974368825898
// '李洵' is not found in the model. // error alert in console
false

w2v.substract(word1 = "word1", word2 = "word2")

w2v.substract(vector1 = [], word2 = "word2")

w2v.substract(word1 = "word1", vector2 = [])

w2v.substract(vector1 = [], vector2 = [])

Substract vector1 from vector2.
Will auto search the vector of passed word in model. Return false if it's not found.

Params	Description	Default Value
word1	Subtrahend	No default value
word2	Minuend	No default value

Example:

'use strict';
var w2v = require("./lib");
var modelFile = "./data/test.model.bin";
w2v.load( modelFile );
var a = w2v.substract("孫悟空", "孫悟空");
console.log(a);

Sample Output:

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

w2v.add(word1 = "word1", word2 = "word2")

w2v.add(vector1 = [], word2 = "word2")

w2v.add(word1 = "word1", vector2 = [])

w2v.add(vector1 = [], vector2 = [])

Add vector1 to vector2.
Will auto search the vector of passed word in model. Return false if it's not found.

Params	Description	Default Value
word1	Summand	No default value
word2	Addend	No default value

Example:

'use strict';
var w2v = require("./lib");
var modelFile = "./data/test.model.bin";
w2v.load( modelFile );
var a = w2v.add("孫悟空", "孫悟空");
var b = w2v.getVector("孫悟空");
console.log(a);
console.log(b);

Sample Output:

[ 0.208812,
  -0.320038,
  -1.209012,
  -1.245608,
  0.078964,
  -0.240116,
  0.14711,
  0.11292,
  0.198118,
  -0.838564 ]
[ 0.104406,
  -0.160019,
  -0.604506,
  -0.622804,
  0.039482,
  -0.120058,
  0.073555,
  0.05646,
  0.099059,
  -0.419282 ]