The easiest way to get an array of n-grams from a string or array of tokens!

No dependencies!

npm install --production --save simplengrams

SimpleNGrams exports one function: nGram().

The function take the following arguments:

• input - a string or array of strings to be split into ngrams.
• n - the ngram size as a number. Defaults to 2 (i.e. bigrams).
• pad - optional padding parameter. Takes a boolean or an array. Defaults to false (i.e. no padding). See Padding below.
• splitPattern - optional pattern as string or RegExp to split input string by. Defaults to spaces. See Pattern below.

import { nGram } from "simplengrams";
const text = "In the beginning God created the heavens and the earth.";
const bigrams = nGram(text);
console.log(bigrams);

Output:

[
  ["In", "the"],
  ["the", "beginning"],
  ["beginning", "God"],
  ["God", "created"],
  ["created", "the"],
  ["the", "heavens"],
  ["heavens", "and"],
  ["and", "the"],
  ["the", "earth"],
  ["earth", "."],
];

Custom padding options can be used to add right and left padding to the output array.

The padding argument is the third argument in nGram(). It takes a boolean (i.e. true or false) or an array.

The padding option defaults to false if it is not supplied.

Some examples:

• false = padding is not applied.

• true = null is used as padding.

• ['FOO', 'BAR'] = The string 'FOO' is used as left padding and the string 'BAR' is used as right padding.

• ['FOOBAR'] = The string 'FOOBAR' is used as both left and right padding.

You can disable individual padding by using undefined like so:

• [undefined, 'BAR'] will disable left padding and use 'BAR' as right padding.

N.B. null will cause the padder to use the null element literally. Use undefined instead to disable padding.

N.B. Simply use false instead of [undefined, undefined] - it results in the same output but is slightly faster.

import { nGram } from "simplengrams";
const text = "In the beginning God created the heavens and the earth.";
const bigrams = nGram(text, 2, true);
console.log(bigrams);

[
  [null, "In"],
  ["In", "the"],
  ["the", "beginning"],
  ["beginning", "God"],
  ["God", "created"],
  ["created", "the"],
  ["the", "heavens"],
  ["heavens", "and"],
  ["and", "the"],
  ["the", "earth"],
  ["earth", "."],
  [".", null],
];

import { nGram } from "simplengrams";
const text = "In the beginning God created the heavens and the earth.";
const bigrams = nGram(text, 2, [undefined, "END"]);
console.log(bigrams);

[
  ["In", "the"],
  ["the", "beginning"],
  ["beginning", "God"],
  ["God", "created"],
  ["created", "the"],
  ["the", "heavens"],
  ["heavens", "and"],
  ["and", "the"],
  ["the", "earth"],
  ["earth", "."],
  [".", "END"],
];

The pattern argument is an optional fourth argument of nGram().

It defaults to /\s+/ and can take a string or a RegExp.

If your input is an array, the pattern argument is ignored.

For string inputs, the pattern argument is used to split the string into tokens, exactly like string.split().

Use node index.test.js to check any development changes against expected outputs.

© 2017-24 P. Hughes. All rights reserved.

Shared under the MIT license license.