Classifying spam using a Transformer Pytorch model. Modern spam filters still use simple Naive Bayesian classifiers. This project hopes to surpass the accuracy of these previous models using a Transformer architecture.

Before you do anything you'll need to install the dependencies in the requirements.txt file. I would recommend setting up a Python virtual environment first to keep this projects required packages separate from those on your system.

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

To train on your own emails rename the .env.example file to .env and fill out the values. Run the email_dataset.py script to create the database file. Finally, run train.py to train the model.

I have another training dataset that I created from all the available training data I could find out on the internet. You can find it at https://github.com/realprogrammersusevim/email-dataset.

Once you have the training dataset and you've run the train.py script you should be left with a model.pth file that you can use to classify and categorize spam and ham email.

Use the model to classify new emails as they come in and automatically move them to Junk if they're identified as junk.