Skip to content
/ nlc Public
forked from stanfordmlgroup/nlc

Neural Language Correction implemented on Tensorflow

Notifications You must be signed in to change notification settings

moberq/nlc

 
 

Repository files navigation

Neural Language Correction

Introduction

Implementation of Neural Language Correction (corresponding paper) on Tensorflow.

Requirements

  • Python 2.7.x (anaconda2)
  • tensorflow-gpu 0.12.1
  • kenlm 0.0.0

Parallel data (for training && validation) crawled from lang-8

datasets download link

P.S. training set: 1,987,139 pairs, validation set: 496,785 pairs.

Pretrained TensorFlow Model

ckpt download link

n-gram Language Model (generated by kenlm)

lm download link

Training (from scratch)

$ python train.py

Interactive decoding

$ python decode.py

About

Neural Language Correction implemented on Tensorflow

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%