node-crf

A node wrapper for the CRF++ toolkit

Attention

This module assume you to have a basic knowledge about CRF, so I suggest you to look at the official site of the original tool ( CRF++)

Requirements

• node-gyp
• A trained CRF++ model

Installation

From the sources

In the folder run

  npm build .

From npm

 npm install node-crf

Usage

// Require the module
var nodecrf = require('node-crf');

// Create a new classifier object
// first argument: path to the model file
// second argument: Nbest value
// third argument: Deep info
// NOTE: up to know the module ignore the deep info parameter
var classifier = new nodecrf.CRF('model', 2, true);

// Specifiy the sperator "WordSEPARATORPostTag"
classifier.separator = '_';

// Initialize the classifier
// Any changes done on the object after the invokation of this method will be ignored
classifier.init();

var text = '2_CD nd_CC part_NN of_IN Web_NNP User_NN Engagement_NN tutorial:_NN Web_NNP Analytics_NNS is_VBZ about_IN to_TO start_VB by_IN @mounialalmas_NNS #www_NN 2013_CD';

// Classify the piece of text
// It will return an array with the selected (n best) categories for each word
var solution = classifier.classify(text);

console.log(solution);

Known Issues

• You may need to add config.h to deps/crfpp/config directory according to your platform and arch, now we support linux/ia32, linux/x64, mac/x64
• Unresolved compilation issue on Windows (sorry guys, I'm working on that :( )

Acknowledgments

• Thanks to Vimos for adding the 'nbest' feature and for "gypifying" the CRF++ library for the module