Machine Learning to Predict Sarcasm in Article headlines

Abstract

This paper describes a machine learning model designed to assign sets of possible labels to article headlines. This was done by analyzing a dataset with article headlines from 2 sources: The Onion and The Huffpost. The Onion is a site known for generally sarcastic headlines, while the Huffpost has more genuine headlines. The data was then inputted into a Zero-Shot-Classification model, which labeled each headline as either with labels generally pertaining to "fake" or "real”. This accuracy data is subsequently displayed with bar charts.