Detecting hate speech on Twitter.

First, you need to download and configure the data.world package (you can get an API key here):

pip install datadotworld
dw configure

You can now fetch the dataset provided by Davidson et al. (2017):

python fetch_dataset.py

This will create a data directory with three files inside: tweets (full dataset), tweets_train (training set), and tweets_test (test set).

First, install the Python interface for fastText:

pip install cython
pip install fasttext

Train the fastText model:

python model_fasttext.py

This will create a file classifier.bin in the bin/ directory.

Train the Keras bi-LSTM model:

python model_keras.py