Neural Machine Translation

This mini-project explores the task of solving translating between two languages using sequential deep learning models.

For this project, we use a simpler task of translating between date formats. We use a bi-directional LSTM model. We also make use of a soft attention mechanism. This allows the decoder model to learn which parts of the input to concentrate on for translating specific parts of the input.

Visualizing attention weights

For example, here we see that for outputting translated year, the model attends to the year part. For the month, it attends to the month.

This project was completed as part of the Coursera course on Sequential Models.