nlp-js-tools-french

NLP Javascript tools for french language

Tokenize, POS Tagger, lemmatizer and stemmer

This package is partly based on the Snowball stemming algorythm and the javascript adaptation by Kasun Gajasinghe, University of Moratuwa

This package offers 4 NLP tools in javascript for french language :

• Tokenizing
• POS Tagging
• Lemmatizing
• Stemming

Install

npm install nlp-js-tools-french

Usage

var NlpjsTFr = require('nlp-js-tools-french');

Corpus to use

var corpus = "Elle semble se nourrir essentiellement de plancton, et de hotdog.";

Configs

var config = {
    tagTypes: ['art', 'ver', 'nom'],
    strictness: false,
    minimumLength: 3,
    debug: true
};

New instance with specific corpus and configs

var nlpToolsFr = new NlpjsTFr(corpus, config);

These are the available methods, self-explanatory. Note: The sentence that is passed into the class earlier is automaticaly tokenized.

var tokenizedWords = nlpToolsFr.tokenized;
var posTaggedWords = nlpToolsFr.posTagger();
var lemmatizedWords = nlpToolsFr.lemmatizer();
var stemmedWords = nlpToolsFr.stemmer();
var stemmedWord = nlpToolsFr.wordStemmer("aléatoirement");

Attributes

config

Shows config

tokenized

["semble", "nourrir", "de"]

Methods return

posTagger()

[{
  "id": 1,
  "word": "semble",
  "pos": [
   "VER",
   "VER"
  ]
 },
 {
  "id": 2,
  "word": "nourrir",
  "pos": [
   "VER"
  ]
 },
 {
  "id": 3,
  "word": "de",
  "pos": [
   "NOM",
   "ART:def",
   "PRE"
  ]
 }]

lemmatizer()

[{
  "id": 1,
  "word": "semble",
  "lemma": "sembler"
 },
 {
  "id": 2,
  "word": "nourrir",
  "lemma": "nourrir"
 },
 {
  "id": 3,
  "word": "de",
  "lemma": "de"
 }]

stemmer()

[{
  "id": 1,
  "word": "semble",
  "stem": "sembl"
 },
 {
  "id": 3,
  "word": "nourrir",
  "stem": "nourr"
 },
 {
  "id": 5,
  "word": "de",
  "stem": "de"
}]

wordStemmer(word)

{
    word: "aléatoirement",
    stem: "aléatoir"
}

Config

Option	Type	Default	Description
tagTypes	Array	["adj", "adv", "art", "con", "nom", "ono", "pre", "ver", "pro"]	List of dictionnaries the package will look in, in case you only need verbs or nouns, both or whatever else. If a word does not belong to any type, it is tagged as "UNK".
strictness	Bool	false	If you set the strictness to true and try to POS Tag the word generalement, it will fail because the word is missine its accents. On the other hand, trying to POS Tag the word dé with the strictness set to false well return the types art, pre and nom because the word will match de in these dictionnaries.
minimumLength	Int	1	Algorythms will ignore words that are shorter than this parameter.
debug	Bool	false	Enable console debug