ts-spelling-corrector

A lightweight TypeScript spelling corrector inspired by Peter Norvig's algorithm. It allows correcting individual words or full sentences by training on a given corpus.

Features

• Corrects misspelled words using probabilistic matching
• Supports training from plain text or file-based corpora
• Token-preserving sentence correction
• Written in TypeScript and fully typed
• Simple API, zero dependencies

Installation

pnpm add ts-spelling-corrector
# or
npm install ts-spelling-corrector
# or
yarn add ts-spelling-corrector

Quick Start

Basic Usage

import { SpellCorrector } from 'ts-spelling-corrector';

const corrector = new SpellCorrector();

// Add corpus from a string
corrector.addStringCorpus('the quick brown fox jumps over the lazy dog');

console.log(corrector.correctWord('quik'));
// => 'quick'

console.log(corrector.correctSentence('teh quik borwn fx.'));
// => 'the quick brown fox.'

Load from File

await corrector.addFileCorpus('./path/to/corpus.txt');

The corpus should be plain text. The frequency of each word is computed automatically.

API

addStringCorpus(text: string): void

Creates a corpus from a string of words. Words are tokenized and counted.

addFileCorpus(path: string): Promise<void>

Loads and parses a text file to build the corpus.

correctWord(word: string): string

Returns the most likely correction for a given word.

correctSentence(sentence: string): string

Corrects a sentence while preserving punctuation and spacing.

Development

Scripts

| Command | Description | | --------------- | ------------------------------ | | pnpm dev | Run in dev mode with tsx | | pnpm build | Build both ESM and CJS outputs | | pnpm test | Run unit tests with Vitest | | pnpm coverage | Generate coverage report | | pnpm lint | Lint the codebase with Biome |

Code is formatted and linted with Biome, and tested using Vitest.

License

MIT © Valère Pique