This project aims to build end-to-end machine translation pipeline capable of translating English text input to French output.

Different neural network architectures are evaluated and compared and details of their implementation can be found in the report.

1. Create (and activate) a new environment with Python 3.6. ** Linux or Mac: **

conda create --name mtdl python=3.6
source activate mtdl

2. Clone the repository and navigate to root of the repo. Then install several dependencies.

git clone https://github.com/n-lamprou/MachineTranslation.git
cd MachineTranslation
pip install .

3. For using jupyter notebooks, create an IPython kernel for the cvdl environment.

python -m ipykernel install --user --name mtdl --display-name "mtdl"

To train a translator, the learn.py script needs to be executed. The additional -net arguement is usedd to choose the network architecture to be used. An example is shown below and the flag -h can be used to display all options.

python learn.py -net EmbeddingRNN

To translate some text run the translate.py script, followed by the network architecture of choice using the -net flag. In the terminal window, type in the English phrase you need to trannslate and hit return.

python translate.py -net EmbeddingRNN