QANTA

From QBWiki
Jump to navigation Jump to search

QANTA (Question Answering Neural network with Trans-sentential Averaging) is a computerized question-answering system that was designed to answer quizbowl tossups using a bank of quizbowl questions and answer-lines from the last twenty years. What differentiated itself from other computer playing machines (such as Watson) is the fact that it answers information when its confidence in knowing the answer is high enough, rather than seeing the entire question at once before determining if it knows it.

The QANTA project is headed up by Jordan Boyd-Graber, a professor at the University of Colorado, Boulder.

It played an exhibition match during HSNCT against a group that consisted of Colby Burnett, Ben Ingram, Alex Jacob, and Kristin Sausville representing "Humanity". Using HSNCT tossups from round 15 and 16, QANTA and Humanity tied with a scoreline of 200-200.

On 2 October 2015, at the University of Washington, QANTA was pitted against Ken Jennings in an exhibition match. QANTA won, 300-160.

At the 2016 HSNCT there was a rematch, with Team Humanity now consisting of Rob Carson, Scott Blish, Dylan Minarik, and Niki Peters. This time, Humanity took no chances, and defeated QANTA, 345-145.


External Links