I trained a model on headlines and Facebook API reaction data, to be able to predict the stressfulness of news stories, using scikit-learn and NLTK.
The purpose of the app is to assist people who like to read daily news with conscious consumption of stressful news stories. I think that starting your day with MyChillNews is one of those things that feels good, and is also good for you.
The codebase consists of two parts:
modelThis code relates to "behind the scenes" operations: downloading Facebook data, training the model, and updating the database with today's headlines and their predicted stress impacts.
appThis is a Flask application, using Bootstrap, which accesses the database to render the MyChillNews app. It is normally running on an EC2 instance, the server being gunicorn managed by supervisord. Most of the logic is contained in