Detecting fake news using DistilBERT

Detection is based on collection of Twitter posts which contain rumours or non-rumours (fake & real) information.

In addition to analyzing plain text of each tweet graph data (social interactions) is taken into consideration, e.g. retweet count, shares count, etc.

Setup

1. Create venv with your favourite tool
2. Activate it
3. Run

   python install.py

4. Provide dataset.key file in raw directory
5. In raw directory run

bash prepare_dataset.sh  # This will initialize raw dataset

6. Run

python setup_dataset.py  # This will create dataset.csv

Credits & acknowledgements

Original dataset (PHEME) belongs to Elena Kochkina, Maria Liakata & Arkaitz Zubiaga.
It was downloaded from here and then encrypted because of sensitive data inside.