Last week, a computer system named Watson bested two human champions in a practice round of the trivia game Jeopardy!

Computers have beaten humans in chess before, but question answering is even more difficult because it involves the use of natural language. How can a computer figure out the meaning of a question, and pull the right answer out of a knowlege database?

UWA linguist Daniel Midgley might have the right answer on this week’s Talk the Talk.


Listen to this episode

Or subscribe via iTunes
Click here for more options: How to listen to Talk the Talk


Show notes

News articles on Watson:
http://blogs.discovermagazine.com/80beats/2011/01/13/ibms-jeopardy-playing-computer-tromps-human-players-in-practice-round/
http://mashable.com/2011/01/13/watson-jeopardy/

A promo video here:
http://video.nytimes.com/video/2010/06/15/magazine/1247468055784/how-does-watson-work.html

A scholarly paper (PDF) on the state of QA as of 2001, but this precedes the data-driven explosion after that time.
http://www.loria.fr/~gardent/applicationsTAL/papers/jnle-qa.pdf

And you can try beating Watson yourself.
http://www.nytimes.com/interactive/2010/06/16/magazine/watson-trivia-game.html
I wonder if all the human players would beat Watson if their answers were aggregated. I bet they would.