Learn about our RFC process, Open RFC meetings & more.Join in the discussion! »

@simonjb/rake-js

0.1.2 • Public • Published

RAKE.js

A pure JS implementation of the Rapid Automated Keyword Extraction (RAKE) algorithm. Put in any text corpus, get back a bunch of keyphrases and keywords.

TypeScript Build Status styled with prettier License: LGPL v3

Currently supported languages:

  • english
  • german
  • spanish
  • italian
  • dutch
  • portugese
  • swedish

More languages are fairly easy to add, see the stoplist module for details.

How to use

Without any further options:

  import rake from 'rake-js'
 
  const myKeywords = rake(someTextContent) // ['keyword1, ...]

When the language is known in advance (faster execution):

  import rake from 'rake-js'
 
  const myKeywords = rake(someTextContent, { language: 'english' })

When the corpus is divided by something other than whitespace (eg: ;):

  import rake from 'rake-js'
 
  const myKeywords = rake(someTextContent, { delimiters: [';+'] })

Implementation Details

This algorithm is fast, compared with other approaches like TextRank. The results are surprisingly good for a cross-language algorithm, and the truly relevant keywords / phrases are included in the result in most cases. For more details about the RAKE algorithm, read the original paper.

There are still rough edges in the code, but I tried to translate the abstract algorithm into a solid software package, tested and typesafe. Actually I wrote this thing because I was very disappointed with all the existing solutions on NPM, and I hope this repository is easier to contribute to in the future.

Roadmap:

  • support more languages (only handful are whitelisted for now)
  • duplicate keyword filtering
  • check browser compatibility

LICENSE:

LGPL-3.0.

You can use this package in all your free or commercial products without any issues, but I want bugfixes and improvements to this algorithm to flow back into the public code repository.

Install

npm i @simonjb/rake-js

DownloadsWeekly Downloads

2

Version

0.1.2

License

LGPL-3.0

Unpacked Size

272 kB

Total Files

92

Last publish

Collaborators

  • avatar