Neural-Machine-Translation

Running the models:

• Run train_and_test.ipynb to train and test all models
• To use them on command line, do:
  • jupyter nbconvert --to=script train_and_test.ipynb
  • ipython train_and_test.py

Architecture Diagrams:

• seq2seq: Sequence to Sequence Learning with Neural Networks
• attention: Neural Machine Translation By Jointly Learning To Align And Translate
• effective approaches: Effective Approaches to Attention-based Neural Machine Translation
• coverage: Modeling Coverage for Neural Machine Translation

Results obtained on the sample dataset (english -> tamil):

Click on the model name to download trained models

Model (Seq2Seq)	Bleu-Score
Linguistic coverage	0.089
General attention	0.087
Fertility coverage	0.0829
Vanilla	0.082
Concat attention	0.0814
Dot attention	0.079
MLP attention	0.072