Skip to content

🍰 A library for creating n-grams, skip-grams, bag of words, bag of n-grams, bag of skip-grams.

License

Notifications You must be signed in to change notification settings

DanielJohnBenton/ngrams.js

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

31 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

ngrams.js

A Node.js library for creating n-grams, skip-grams, bag of words, bag of n-grams, bag of skip-grams. Also has a simple function to sanitise input text and split it into words for use in these methods.

Usage

Put ngrams.js in your node_modules folder and require it:

let __ngrams = require("ngrams");

Input

These methods take an array of words to turn into n-grams, skip-grams, etc.

"use strict";
let __ngrams = require("ngrams");

// Create 2-grams
let ngrams = __ngrams.Ngrams(["these", "are", "some", "words"], 2);
console.log(ngrams);

Output:

[ 'these are', 'are some', 'some words' ]

"these are some words".split(" ") is nice and simple, but often our source text is not so simple. You may choose to use some sort of NLP tokeniser. This library provides a rudimentary method, SanitiseToWords, to get rid of things like double spaces, punctuation, and so on in a string before splitting it into words.

console.log(__ngrams.SanitiseToWords("   Turning and turning in the widening gyre\r\n    The falcon cannot hear the falconer;\r\n    Things fall apart; the centre cannot hold;\r\n    Mere anarchy is loosed upon the world   "));

Output:

[ 'Turning', 'and', 'turning', 'in', 'the', 'widening', 'gyre', 'The', 'falcon', 'cannot', 'hear', 'the', 'falconer', 'Things', 'fall', 'apart', 'the', 'centre', 'cannot', 'hold', 'Mere', 'anarchy', 'is', 'loosed', 'upon', 'the', 'world' ]

See documentation below for more information.

Methods

🐚 Ngrams

Create n-grams from an array of words.

Parameter Type Description
words ARRAY [INTEGER] = STRING An array of words e.g. ["these", "are", "words"]
n INTEGER Size of the n-grams, e.g. 2 will create bigrams ["these are", "are words"]

Returns an array of n-grams of size n (ARRAY [INTEGER] = STRING).

let words = __ngrams.SanitiseToWords("   Turning and turning in the widening gyre\r\n    The falcon cannot hear the falconer;\r\n    Things fall apart; the centre cannot hold;\r\n    Mere anarchy is loosed upon the world   ");
let ngrams = __ngrams.Ngrams(words, 5);
console.log(ngrams);

Output (truncated):

[ 'Turning and turning in the',
  'and turning in the widening',
  'turning in the widening gyre',
...

🐚 SkipGrams

Create skip-grams from an array of words.

Parameter Type Description
words ARRAY [INTEGER] = STRING An array of words e.g. ["these", "are", "words"]
size INTEGER Size of the n-grams e.g. 2: "these are", "are words"
distance INTEGER Distance to skip to create skip-grams, e.g. 5 will create skip-grams using the base word (or n-gram) and n-grams from the 5 following words.
sortForDuplicates INTEGER Pass __ngrams.SORT_NGRAMS or __ngrams.DONT_SORT_NGRAMS. Sorting n-grams alphabetically can help flag up duplicates e.g. when creating a bag of words/n-grams/skip-grams. If you only care about pairing n-grams by proximity but not by direction, use __ngrams.DONT_SORT_NGRAMS.

Returns an array of n-grams found near one another within distance words (ARRAY [INTEGER][INTEGER] = STRING).

let words = __ngrams.SanitiseToWords("   Turning and turning in the widening gyre\r\n    The falcon cannot hear the falconer;\r\n    Things fall apart; the centre cannot hold;\r\n    Mere anarchy is loosed upon the world   ");

let skipGrams = __ngrams.SkipGrams(words, 1, 2, __ngrams.DONT_SORT_NGRAMS);
console.log(skipGrams);

Output (truncated):

[ [ 'Turning', 'and' ],
  [ 'Turning', 'turning' ],
  [ 'and', 'turning' ],
  [ 'and', 'in' ],
  [ 'turning', 'in' ],
  [ 'turning', 'the' ],
...

You can choose instead to pass __ngrams.SORT_NGRAMS and this will make direction irrelevant (e.g. it will be easier to sport ["Turning", "and"] and ["and", "turning"] as the same words because they are now sorted to ["Turning", "and"] and ["turning", "and"]. Using method BagOfSkipGrams (passing __ngrams.CASE_INSENSITIVE) would then remove one of these as a duplicate.

let skipGrams = __ngrams.SkipGrams(words, 1, 2, __ngrams.SORT_NGRAMS);
console.log(skipGrams);

Output (truncated):

[ [ 'and', 'Turning' ],
  [ 'turning', 'Turning' ],
  [ 'and', 'turning' ],
  [ 'and', 'in' ],
  [ 'in', 'turning' ],
  [ 'the', 'turning' ],
...

🐚 BagOfNgrams

Generates n-grams and removes duplicates. Can be case sensitive or insensitive by passing __ngrams.CASE_SENSITIVE or __ngrams.CASE_INSENSITIVE.

Parameter Type Description
words ARRAY [INTEGER] = STRING An array of words e.g. ["these", "are", "words"].
n INTEGER Size of the n-grams e.g. 2 creates bigrams ["these are", "are words"]
caseSensitivity INTEGER Pass __ngrams.CASE_SENSITIVE or __ngrams.CASE_INSENSITIVE. Case insensitive calls will ignore differences in case when removing duplicates e.g. "Turning", "turning", "TURNING" will all be seen as identical and reduced to just "Turning".

Returns an array of n-grams with duplicates removed (ARRAY [INTEGER] = STRING).

let words = __ngrams.SanitiseToWords("   Turning and turning in the widening gyre\r\n    The falcon cannot hear the falconer;\r\n    Things fall apart; the centre cannot hold;\r\n    Mere anarchy is loosed upon the world   ");

let bagOfWords = __ngrams.BagOfNgrams(words, 1, __ngrams.CASE_INSENSITIVE);
console.log(bagOfWords);

Output:

[ 'Turning',
  'and',
  'in',
  'the',
  'widening',
  'gyre',
  'falcon',
  'cannot',
  'hear',
  'falconer',
  'Things',
  'fall',
  'apart',
  'centre',
  'hold',
  'Mere',
  'anarchy',
  'is',
  'loosed',
  'upon',
  'world' ]

🐚 BagOfWords

This is just a wrapper function for readability that calls BagOfNgrams with an n-gram size (n) of 1.

Parameter Type Description
words ARRAY [INTEGER] = STRING An array of words e.g. ["these", "are", "words"].
caseSensitivity INTEGER Pass __ngrams.CASE_SENSITIVE or __ngrams.CASE_INSENSITIVE. Case insensitive calls will ignore differences in case when removing duplicates e.g. "Turning", "turning", "TURNING" will all be seen as identical and reduces to just "Turning".

Returns an array of words with duplicates removed (ARRAY [INTEGER] = STRING).

let bagOfWords = __ngrams.BagOfWords(words, __ngrams.CASE_INSENSITIVE);

🐚 BagOfSkipGrams

Generates skip-grams and removes duplicates. Can ignore direction by passing __ngrams.SORT_NGRAMS. Can be case insensitive by passing __ngrams.CASE_INSENSITIVE.

Parameter Type Description
words ARRAY [INTEGER] = STRING An array of words e.g. ["these", "are", "words"]
size INTEGER Size of the n-grams e.g. 2: "these are", "are words"
distance INTEGER Distance to skip to create skip-grams, e.g. 5 will create skip-grams using the base word (or n-gram) and n-grams from the 5 following words.
sortForDuplicates INTEGER Pass __ngrams.SORT_NGRAMS or __ngrams.DONT_SORT_NGRAMS. Sorting n-grams alphabetically can help flag up duplicates e.g. when creating a bag of words/n-grams/skip-grams. If you only care about pairing n-grams by proximity but not by direction, use __ngrams.DONT_SORT_NGRAMS.
caseSensitivity INTEGER Pass __ngrams.CASE_SENSITIVE or __ngrams.CASE_INSENSITIVE. Case insensitive calls will ignore differences in case when removing duplicates e.g. "Turning", "turning", "TURNING" will all be seen as identical and reduces to just "Turning".

Returns an array of arrays of paired n-grams ARRAY [INTEGER][INTEGER] = STRING.

Case sensitive, direction sensitive:

let words = "Something and SOMETHING and something and something".split(" ");

let skipGrams = __ngrams.BagOfSkipGrams(words, 2, 2, __ngrams.DONT_SORT_NGRAMS, __ngrams.CASE_SENSITIVE);
console.log(skipGrams);

Output:

[ [ 'Something and', 'and SOMETHING' ],
  [ 'Something and', 'SOMETHING and' ],
  [ 'and SOMETHING', 'SOMETHING and' ],
  [ 'and SOMETHING', 'and something' ],
  [ 'SOMETHING and', 'and something' ],
  [ 'SOMETHING and', 'something and' ],
  [ 'and something', 'something and' ],
  [ 'and something', 'and something' ],
  [ 'something and', 'and something' ] ]

Case sensitive, direction insensitive __ngrams.BagOfSkipGrams(words, 2, 2, __ngrams.SORT_NGRAMS, __ngrams.CASE_SENSITIVE):

[ [ 'and SOMETHING', 'Something and' ],
  [ 'Something and', 'SOMETHING and' ],
  [ 'and SOMETHING', 'SOMETHING and' ],
  [ 'and something', 'and SOMETHING' ],
  [ 'and something', 'SOMETHING and' ],
  [ 'something and', 'SOMETHING and' ],
  [ 'and something', 'something and' ],
  [ 'and something', 'and something' ] ]

Case insensitive, direction insensitive __ngrams.BagOfSkipGrams(words, 2, 2, __ngrams.SORT_NGRAMS, __ngrams.CASE_INSENSITIVE):

[ [ 'and SOMETHING', 'Something and' ],
  [ 'Something and', 'SOMETHING and' ],
  [ 'and something', 'and SOMETHING' ] ]

Case insensitive, direction sensitive __ngrams.BagOfSkipGrams(words, 2, 2, __ngrams.DONT_SORT_NGRAMS, __ngrams.CASE_INSENSITIVE):

[ [ 'Something and', 'and SOMETHING' ],
  [ 'Something and', 'SOMETHING and' ],
  [ 'and SOMETHING', 'SOMETHING and' ],
  [ 'and SOMETHING', 'and something' ] ]

🐚 ConcatSkipGrams

Pass skip-grams through this method if you would prefer a simpler array of strings where skip-grams have been concatenated into a single string.

Parameter Type Description
skipGrams ARRAY [INTEGER][INTEGER] = STRING An array of skip-grams created using SkipGrams or BagOfSkipGrams which you want to simplify into ARRAY [INTEGER] = STRING by joining each n-gram pair into one string.

Example:

let skipGrams = __ngrams.SkipGrams(["these", "are", "some", "words"], 2, 2, __ngrams.DONT_SORT_NGRAMS);

skipGrams = __ngrams.ConcatSkipGrams(skipGrams);
console.log(skipGrams);

Output:

[ 'these are are some',
  'these are some words',
  'are some some words' ]

🐚 SanitiseToWords

A rudimentary method that attempts to refine messy text into an array of words.

Parameter Type Description
text STRING The source text you want to split into words.

Note that this is mainly only good for English-language text - it does not support accented characters etc.

Its approach is to replace anything outwith a small list of allowable characters with a space, avoiding any double spacing, and then split by those spaces.

This works quite well for many English-language texts - with the occasional mistake.

However, you may prefer to roll your own sanitisation/splitting/tokenisation method based more closely on your source text(s). You might like to use this 'natural' library's tokenizers.

let words = __ngrams.SanitiseToWords("   Turning and turning in the widening gyre\r\n    The falcon cannot hear the falconer;\r\n    Things fall apart; the centre cannot hold;\r\n    Mere anarchy is loosed upon the world   ");
console.log(words);

Output:

[ 'Turning', 'and', 'turning', 'in', 'the', 'widening', 'gyre', 
	'The', 'falcon', 'cannot', 'hear', 'the', 'falconer', 
	'Things', 'fall', 'apart', 'the', 'centre', 'cannot', 'hold', 
	'Mere', 'anarchy', 'is', 'loosed', 'upon', 'the', 'world' ]

Interesting/Useful Links

Acknowledgements

Releases

No releases published

Packages

No packages published