Neural Machine Translation (NMT)

This is a repository for a deep learning model using encoder-decoder architecture to translate german sequences to English automatically.

prerequisites:

• python 3
• Keras 2.0 or higher
• NumPy
• matplotlib

All packages used here can be installed by pip as following the example for NumPy:

pip install numpy

Dataset:

The project downloads the dataset from https://raw.githubusercontent.com/jbrownlee/Datasets/master/deu.txt; you don't need to download it separately.

process, train, test

data_preparation.py downloads the dataset and separates German and English and pre-processes the dataset. load_data.py fits a german and English tokenizer and prepared training, validation, and test data. model.py defines the model and trains the model with german and English sequences training data. evaluate.py evaluates the model on train and test dataset. The BLEU score is the evaluation metric in this project.

Generate translation

python3 model.py