closewords

A library for finding the most similar word from a list of words, supporting Japanese (including kanji).

Note

It may be a little slow because it uses morphological analysis. By adopting worker_threads, the processing speed is slightly improved compared to the standard.

日本語のドキュメントは README-ja.md を参照してください。

Installation

npm install closewords

Usage

closeWords(word, candidates, raw?)

Parameter	Type	Default	Description
word	string | { word: string; pronounce?: string }	—	The reference word or object.
candidates	Array<string | { word: string; pronounce?: string }>	—	List of candidate words or objects.
raw	boolean	false	Whether to include similarity scores.

Returns: Promise<string[]> when raw is false, Promise<Array<{ word: string; score: number }>> when raw is true.

The highest score is 1 (perfect match), the lowest is 0.

The optional pronounce field accepts alphabetic strings only and is used as a pronunciation hint for morphological analysis.

Example

import { closeWords } from 'closewords';

const word = '東京';
const candidates = ['東京', 'とっこう', '東きょう', 'とう京', 'とうきょう', 'とーきょー'];

const result = await closeWords(word, candidates);
console.log(result);
// => [ '東京' ]

const resultWithScores = await closeWords(word, candidates, true);
console.log(resultWithScores);
// => [
//   { word: '東京', score: 1 },
//   { word: 'とう京', score: 0.693... },
//   ...
// ]

Change Log

2.3.1 --> 3.0.0

Rewrote in TypeScript. Introduced tsdown for bundling.
Moved source files to src/. Separated README into English / Japanese.

2.2.0 --> 2.3.0

Improved Hepburn-style completion. Format code.

2.1.4 --> 2.2.0

word.pronounce and pronounce in candidates[] are completed Hepburn-style.

2.1.3 --> 2.1.4

Fixed bugs.

2.1.2 --> 2.1.3

Fixed score calculation.

2.1.1 --> 2.1.2

Fixed score calculation.

2.1.0 --> 2.1.1

Fixed README. Fixed the issue that only a string could be specified in word. Fixed the issue that word.pronounce was ignored. Fixed the issue that non-alphabet could be specified for word.pronounce and pronounce in candidates[]. word.pronounce and pronounce in candidates[] are now optional. Fixed a few other bugs.

2.0.0 --> 2.1.0

Added a way to specify the pronunciation of words.

1.0.2 --> 2.0.0

Introduced fast-levenshtein and fixed score calculation. The similarity of the original strings is also evaluated.

1.0.1 --> 1.0.2

Introduced jaro-winkler and optimized.

1.0.0 --> 1.0.1

Fixed score calculation.

0.x --> 1.0.0

Package released! Introducing morphological analysis.