QANTA:

QANTA Project

The QANTA project is a University of Maryland research project that focuses on using question answering as a platform for research in machine learning and natural language processing. We build computer systems that be fairly compared against each other and expert humans based on a trivia game called quiz bowl. You can see videos of our previous events against top trivia champs or read about our system:

• Associate Professor Jordan Boyd-Graber to appear on Jeopardy on September 26th (2018)
• What Is … a Research-Inspired Route to “Jeopardy!”? (2018)
• UMD Computerized System Beats Human Quiz Bowl Team at Atlanta Exhibition (2017)
• Automatic Translation and t-SNE (2016)
• Question-Answering System Built by UMD, UC Boulder Bests Ken Jennings (2015)

QANTA Dataset

With the cooperation of the quiz bowl community, we compile datasets of questions that can challenge question answering systems. If you're a computer science researcher, this is likely what you came here for!

Human vs. Machine Competitions: Try it!

Think you could do better than IBM Watson in making a trivia-playing computer program? Download, modify, improve, and submit a system for our December 15 competition! Our system is designed to be easily extended.

Want to take on the best computer programs at trivia? Register for our December 15 competition!

Question Writing Challenge

Given our past events, you might think that computers are pretty smart, right You’d be wrong. quiz bowl questions are supposed to be written pyramidally: harder clues at the start, easier clues at the end. The problem is that question writers don't write questions pyramidally for computers: they write questions for humans. If humans could write questions pyramidally for computers, computers would be in a whole heap of trouble, and look pretty dumb. Our goal is to offer a true test for how smart computers are.

Compete in our first QANTA League competition to see if you can outsmart a computer!

Read our preprint about the first round of the competition!