nlp-toolkit

Natural Language Processing Toolkit for node.js

This module covers some basic nlp principles and implementations. The main focus is performance. When we deal with sample or training data in nlp, we quickly run out of memory. Therefore every implementation in this module is written as stream to only hold that data in memory that is currently processed at any step.

Install

npm install nlp-toolkit

Example

Frequency distribution of words in texts. Tokenize, remove stopwords, stem words, count words. Traditionally those steps happen sequentially. But we do not need to tokenize the whole text before removing stopwords.

var nlp = require('nlp-toolkit');
var fs = require('fs');
var es = require('event-stream');
 
fs.createReadStream('./pride_prejudice.txt')
.pipe(es.split())
.pipe(nlp.tokenizer())
.pipe(nlp.stopwords())
.pipe(nlp.stemmer())
.pipe(nlp.frequency())
.on('data', function (freqDist) {
  console.log(freqDist.slice(0, 10));
})
.on('error', function (err) {
  console.error(err);
});

Modules

Tokenizer

.pipe(nlp.tokenizer(options))

options:

attribute	type	description
characters	RegExp	regular expression that describes what characters to strip of off (default /[^\w]/g).
separator	RegExp	regular expression that describes where to split words (default /\s/g).
eliminateNumbers	boolean	discard tokens that only contain numbers (default false).
toLowerCase	boolean	transform every token to lower case (default true).
emptyStrings	boolean	keep empty string when through some previous steps tokens result in length === 0 (default false).

Tokenizer also work in a non-stream context:

var tokens = nlp.tokenizer(string, options);

Stopwords

.pipe(nlp.stopwords(options))

options:

attribute	type	description
defaultLang	string	default language if processed object does not provide a lang attribute (default en).
additionalWords	object	add additional stopwords to the list of stopwords

additionalWords:

attribute	type	description
all	array	list of stopwords to add to every language
default	array	list of stopwords if language is not supported
lang	array	list of stopwords specific to lang

Supported languages: da, de, en, es, fi, fr, hu, it, nl, no, pt, ro, ru, se, tr.

Stopwords also work in a non-stream context:

nlp.stopwords(sentence, options)
.then(function (tokens) {}})
.catch(function (err) { console.error(err); });

Stemmer

.pipe(nlp.stemmer(options))

options:

attribute	type	description
defaultStemmer	string	default stemmer for language if processed object does not provide a lang attribute (default en).

Supported languages: da, de, en, es, fi, fr, hu, it, nl, no, pt, ro, ru, se, tr.

Stopwords also work in a non-stream context:

var tokens = nlp.stemmer(sentence, options);

This module uses the stemmer implementation of Snowball-Stemmer.

Frequency Distribution

.pipe(nlp.frequency())