An LSTM model that reads surnames as a series of characters and outputs a prediction as to which language the name originates from. There are 18 languages of origin total. This is heavily based on this Pytorch tutorial on RNN's. I've changed the vanilla RNN model they use with an LSTM model.

The files lstm.py and plot.py contain the lstm model and the function that plots the confusion matrix. Just run the file main.py to train the model and output the plots of loss over time and the confusion matrix like below