This project include five approaches to offensive and hate speech detection. Two of which are feature base methods (Bag-of-words model, feature model).BoW and syntactic features are extracted and logistic regression is applied. Other three are neural networks (BERT model, attention model, Character CNN). Each module contains of configuration file (config.jsn), model implementation and run script for training/testing (main_{model}.py).
You need to install the packages from requirements.txt. PyTorch and some others to be installed differently. See their documentation for more information.