Attention-based NMT with a coverage mechanism to indicate whether a source word is translated or not
Python Prolog
Switch branches/tags
Nothing to show
Clone or download
Latest commit 2d40f83 Mar 8, 2018
Permalink
Failed to load latest commit information.
build/lib cleaning the code May 27, 2016
dist cleaning the code May 27, 2016
doc cleaning the code May 27, 2016
experiments Update README.md Nov 10, 2016
groundhog.egg-info cleaning the code May 27, 2016
groundhog cleaning the code May 27, 2016
scripts cleaning the code May 27, 2016
tutorials cleaning the code May 27, 2016
LICENSE cleaning the code May 27, 2016
README.md Update README.md Mar 8, 2018
setup.py cleaning the code May 27, 2016

README.md

NMT-Coverage

Please refer to a clean version of improved NMT, which is incorporates coverage, context gates, and reconstruction models.

In this version, we introduce a coverage mechanism (NMT-Coverage) to indicate whether a source word is translated or not, which proves to alleviate over-translation and under-translation. If you use the code, please cite our paper:

@InProceedings{Tu:2016:ACL,
  author    = {Tu, Zhaopeng and Lu, Zhengdong and Liu, Yang and Liu, Xiaohua and Li, Hang},
  title     = {Modeling Coverage for Neural Machine Translation},
  booktitle = {Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics},
  year      = {2016},
}

For any comments or questions, please email the first author.

Installation

NMT-Coverage is developed by Zhaopeng Tu, which is on top of lisa-groudhog (https://github.com/lisa-groundhog/GroundHog). It requires Theano0.8 or above version (for the module "scan" used in the trainer).

To install NMT-Coverage in a multi-user setting

python setup.py develop --user

For general installation, simply use

python setup.py develop

NOTE: This will install the development version of Theano, if Theano is not currently installed.

How to Run?

See experiments/nmt/README.md