Kyori

A library that offers many methods to calculate string edit distance. In addition, it offers the Kyori algorithm, a simple-yet-nice approach to rank several results against a set of terms (keywords). It is handy when you have search results, and you want to return the most relevant lexicographically.

In Japanese, the word for “distance” is 距離 pronounced as kyori (きょり).

Install

Installing the library is simple, just run the following command.

npm install kyori

Run unit tests

To execute locally the testing suite, just execute:

npm test

Implemented methods

It provides support for the following methods for edit string distance:

• Levenshtein: It measures the minimum number of single-character edits—insertions, deletions, or substitutions—required to change one word into the other.
• Damerau-Levenshtein: An extension of the Levenshtein distance, this method also accounts for transpositions (swapping of two adjacent characters) in addition to insertions, deletions, and substitutions. It is particularly useful when transposition errors are common, such as in typographical mistakes.
• Hamming Distance: This method calculates the number of positions at which the corresponding characters in two strings of equal length are different. It is only applicable when the strings are of the same length and is often used in error detection and correction algorithms.

... and for string similarity:

• Jaro-Winkler: This method is particularly effective for short strings such as names. It calculates a similarity score based on the number and order of common characters, giving higher scores to strings that match from the beginning.
• Kyori: A similarity method that is token sensitive and focus of the similarity of a term to a specific text. It is ideal to rank results for autocomplete interfaces.

Except, kyori, hamming. Jaro-Winkler and levensthein all the other methods are wrapper on the talisman library ↗. For these early versions, this library will be used as a dependency, but since it has not been updated for 3-4 years, it will be replaced in future versions.

Usage

Using the library is also straightforward. First, you must include the top-level kyori module,

const kyori = require('kyori');

the object is structured as followed:

{
    kyori: {
        levensthein: distance(),
        damerau_levensthein: distance(),
        hamming: distance(),
        jaro_winkler: similarity(),
        kyori: similarity()
    }
}

and then call the appropriate method you want to use. You can also include the desired method directly.

Levensthein

Definition can be found here ↗. This metric is natively implemented.

const { levensthein } = require('kyori');

Example:

levensthein.distance('foo', 'foo')  // should be 0
levensthein.distance('foo', 'food') // should be 1
levensthein.distance('foo', null)   // should be -1

Damerau-Levensthein

Definition can be found here ↗.

const { damerau_levensthein } = require('kyori');

Example:

damerau_levensthein.distance('foo', 'foo');   // should be 0
damerau_levensthein.distance('foo', 'food');  // should be 1
damerau_levensthein.distance('foo', 'foodo'); // should be 2

Hamming distance

Definition can be found here ↗. This metric is natively implemented.

const { hamming } = require('kyori');

Usage:

Keep in mind that hamming distance works only for string with same length.

hamming.distance('foo', 'foo')  // should be 0
hamming.distance('foo', 'fob')  // should be 1
hamming.distance('foo', 'food') // should be -1

Jaro-Winkler

Definition can be found here ↗.

const { jaro_winkler } = require('kyori');

Usage:

jaro_winkler.similarity('foo', 'foo') // should be 1
jaro_winkler.similarity('foo', 'bar') // should be 0

Kyori

const { kyori } = require('kyori');

Usage:

kyori.similarity('foo', 'foo')  // should be 0
kyori.similarity('foo', 'food') // should be 1
kyori.similarity('foo', 'ifoo') // should be 2

License and usage

Released under MIT license and maintained by wizhut.tech ↗.