Automated Coding of Political Campaign Television Advertisements

• In Event: Machine Learning: Text and Images as Data

Brief Overview

We propose to automate the coding of campaign TV advertisements, a traditionally manual task. We use machine learning methods to extract audio and image features from video files and show that machine coding is at least as accurate as human coding.

Authors

• Wonjoon Hwang, Harvard University

• Kosuke Imai, Harvard University

• Alexander Tarr, Princeton University