Overview

This is a final project for our Deep Learning class with Professor Christopher Monterola. In this project, we use the deep learning model `GPT-2` from the OpenAI team and retrain this powerful generative text transformer on news datasets collected from the top news sites in the United States during the period of the 2016 US Presidential Elections. The goal of this model was to create a generative text model that learns the writing patterns and stances of the different news outlets specifically with regard to the top 2 Presidential candidates, Donald Trump and Hillary Clinton, in an effort to uncover the underlying media biases of each news outlet. We explored this topic to see the effect of supposed "fake news" by certain news outlets on the overall image of each candidate.

Report

The presentation can be viewed here: pdf

Acknowledgements

This project was done by my group composed of Carlo Cordero, Criselle David, Gemille Gloria, and myself.