544-smt-project

##Group #51 Neural Machine Translation

This is a Keras implementation of RNN Encoder-Decoder, written for as a final project for CSCI-544.

We built our model from basic Keras and Tensorflow building blocks, without using any pre-written sequence to sequence library. Our code is organized as follows:

1. translate.py is the entry point and takes many command line options.
2. data_feeder.py contains a DataFeeder class, a self sufficient class that handles processing, reading and producing batches from the data.
3. callbacks.py contains callback functions that are used in training sessions.
4. beam_graph.py, beam_search.py, tester.py and testing.py are related to the testing model and searching algorithms used in decoding.

Group members (in alphabetic order):

Wilka Carvalho
Yi Ren
Shuyang Sheng
Yaning Yu

For questions regarding the project, please leave comments or pull requests to:

https://github.com/wcarvalho/544-smt-project