Fall 2016: Ballot Ready / Machine Learning for the Ballot Booth

Cognitive computing is going to help with civic duty.

What’s ballot ready trying to solve?

We believe elected officials matter, from the top of the ballot to the bottom. We created BallotReady to make it easy for every voter to research the candidates on his or her ballot and make informed decisions based on what he or she cares about.BallotReady provides free, nonpartisan background information on every candidate and referendum on a voter’s ballot.

In order for Ballot Ready to prepare candidate ballots, the process involves a team to:

  1. Gather candidate and office names
  2. Find campaign websites
  3. Find relevant page text
  4. Pull issue stance sentences
  5. Edit issue stance sentences
  6. Classify Edited Issue Stances

What are we going to build?

In this fall’s cohort, the app challenge will figure out how strongly IBM’s BlueMix Technology can be applied.  In the 2016 Cognitive Computing Challenge, we figured out how to help classify edited issue stances.  We’ll test to see how far BlueMIx can do the work reliably.

How does the App Challenge help with this great idea?

IBM staff member Frank Greco and Cornelia Bailey are continuing to apply IBM’s BlueMix Technology (including Watson) and other tools to Ballot Ready's manual workflow of gathering high-quality content for voter education. In the 2016 Cognitive Computing Challenge, we figured out how to systematically help with categorizing candidate stances (e.g. Education, Environment, etc). In this round of work, we applied the same and new machine learning tools to figure out if a given webpage contains meaningful candidate stances or endorsements.

The Ballot Ready team presented on this work in February 2017 as part of the Polsky Center's UChicago Technology Summit.