rake-js

TL;DR

Quick Start? Follow the installation and sample usage. Don't forget to read the details before deploying to production.

$3

``bash
npm install @shopping24/rake-js
`

$3

`javascript
const deDict = require('@shopping24/rake-js/dist/de');
const rakejs = require('@shopping24/rake-js');

// From https://de.wikipedia.org/wiki/Jaguar, available under the Creative Commons Attribution-ShareAlike License.
const text = 'Die Kopf-Rumpf-Länge des Jaguars beträgt 112 cm bis 185 cm, hinzu kommt ein 45–75 cm langer Schwanz. ' +
'Die Schulterhöhe liegt im Durchschnitt bei etwa 70 cm. Obwohl insgesamt kräftiger und massiger gebaut als der ' +
'Leopard, ist sein Schwanz deutlich kürzer als der des afrikanisch-asiatischen Verwandten. Das Körpergewicht variiert ' +
'stark zwischen unterschiedlichen Regionen und schwankt zwischen 36 und 158 kg. Weibchen sind dabei etwa 10–20 % ' +
'kleiner und entsprechend leichter als männliche Tiere. Darüber hinaus besteht eine ausgeprägte geographische ' +
'Variation. So sind Jaguare in Nord- und Mittelamerika deutlich kleiner als Jaguare in Südamerika. Männliche Tiere in ' +
'Belize haben im Schnitt etwa ein Gewicht von 60 kg, während Jaguarmännchen in Venezuela und Brasilien um die 90–100 ' +
'kg wiegen. Weibliche Jaguare in Brasilien wiegen durchschnittlich fast 80 kg.[2]';

const { result } = rakejs.extract(text)
.setOptions({ articles: deDict.articles, stopWords: deDict.stopwords.concat(deDict.articles) })
.pipe(rakejs.extractKeyPhrases)
.pipe(rakejs.extractAdjoinedKeyPhrases)
.pipe(rakejs.keywordLengthFilter)
.pipe(rakejs.distinct)
.pipe(rakejs.scoreWordFrequency)
.pipe(rakejs.sortByScore);

console.log(result);
`

Background story

$3

RAKE is the acronym for Rapid Automated Keyword Extraction. The basic algorithm is described by Stuart Rose, Dave Engel,
Nick Cramer and Wendy Cowley in their paper "Automatic keyword extraction from individual documents"
(©2010, John Wiley & Sons, Ltd, Source click here).

In short RAKE describes splitting a text into fragments by stop words. Stop words are always considered to be irrelevant
to the context. The _RAKEd_ result of Red Zebra and Jaguar would therefore be [Red Zebra, Jaguar].

The score is then calculated by counting the individual words and and creating degrees based on the length of found
fragments.

$3

This repository includes advanced methods in addition to the original RAKE description. Furthermore we added a
functional wrapper as feature for a more flexible way of handling keyword extraction. The process consists of these
steps:

0. Extracting fragments from any given text using various available methods.
0. Score the fragments.
0. Retrieve the end result.

Extraction and scoring functions from any source making use of the Phrases and Phrase classes may be used and executed
in the desired order.

For reference you can find included extraction and scoring functions under src/utils/.
Included helper functions are accessible in src/lib/.

Pipeline Reference

The extract method and phrases class provide a convenient way to apply multiple filtering and scoring methods on a
given text.

$3

A Phrases object is created and returned by the extract(text) function.

#### Methods

Method | Arguments | Returns | Description
--- | --- | --- | ---
pipe | method: Callable Function | | method Receives and must return Phrases object. result must be an array of Phrase objects.
toPhrase | phrase: String,
score: Number (_optional_) | | Helper function, creates and returns new Phrase object from given phrase and score. Pushing into result property of Phrase object needs to be done manually.
setOptions | options: Object | | Provided key/value pairs get merged with stored Phrases options and made available to pipeline methods. See documentation of Extraction and Scoring Functions for available options.

#### Properties

Property | Type | Mutable | Description
--- | --- | --- | ---
result | | Y | Contains result array of the previous pipe function. Pipe functions must not change object properties but reassign the full object with either an empty array or an array of Phrase objects.
original | | N | The input provided with the extract(text) function.
options |