Sign In

Communications of the ACM

ACM TechNews

How New Yorker Cartoons Could Teach Computers to Be Funny

View as: Print Mobile App Share:
A classic New Yorker cartoon.

The New Yorker magazine is using crowdsourced algorithms to find the funniest captions for its cartoons.

Credit: Peter Steiner/The New Yorker magazine

The New Yorker magazine is using crowdsourcing algorithms developed at the University of Wisconsin at Madison to mine a massive volume of cartoon caption submissions to find the funniest captions.

The algorithms can swiftly parse out the optimal choice for many options, and  the data it collects might one day be used to teach a computer to understand humor.

Using the New Yorker's caption voting tool, cartoon editors can evaluate what captions strike them as funny, not funny, or somewhat funny. These evaluations are then fed into computer models called adaptive crowdsourcing algorithms. The algorithms attempt to filter out the least funny captions quickly to get more people to vote on the potential winners.

"We want to drive the crowd to the top funniest captions instead of wasting their time with the ones that aren't that funny," says University of Wisconsin professor Rob Nowak.

The algorithms' developers aim to create an adaptive program that could direct researchers to conduct lab tests that are most likely to yield the needed outcomes. Nowak thinks the algorithms could be applied toward answering key questions in science, engineering, and medicine.

From CNet
View Full Article


Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account