Translate to multiple languages

Subscribe to my Email updates

https://feedburner.google.com/fb/a/mailverify?uri=helgeScherlundelearning
Enjoy what you've read, make sure you subscribe to my Email Updates

Monday, August 10, 2015

And the New Yorker Cartoon Contest Winner Is … a Computer by Dina Bass

Follow on Twitter as @dinabass
Dina Bass writes, "The magazine’s cartoon editor collaborated with Microsoft researchers on an artificial intelligence project that aims to teach machines what’s funny."


A sample New Yorker cartoon, with the associated tags, that was fed into Microsoft's artificial intelligence software.
Source: Microsoft via Bloomberg

Since 2005, the back page of the New Yorker has usually featured a wordless, black-and-white cartoon, and the funniest reader-submitted caption gets published in a following issue. The magazine’s caption contest has become a fan favorite over the last decade, and the cartoon department receives some 5,000 entries each week. This has become an overwhelming number of jokes to sift through—particularly for Bob Mankoff’s assistant. The 71-year-old cartoon editor for the New Yorker says the average tenure of his assistants is barely a couple of years because he keeps burning them out. “The process of looking at 5,000 caption entries a week usually destroys their mind in about two years, and then I get a new one,” Mankoff says. “It's a little bit daunting. It's like going snow blind; you go humor blind.”

Soon, Mankoff’s assistants could get relief in the form of an assistant of their own: an artificial intelligence system with a sense of humor. Mankoff collaborated with researchers at Microsoft on an artificial intelligence project that aims to teach a computer what’s funny. They’re feeding an archive of New Yorker cartoons and caption-contest entries into AI software to give machines some understanding of humor (the New Yorker’s brand of humor, at least). A Microsoft researcher plans to present the findings (PDF) onstage on Aug. 13 at the KDD data conference in Sydney.

The idea for the project arose at a different convention about a year ago. Dafna Shahaf, a researcher at Microsoft, attended a speech by Mankoff about the cartoon archive, and she left feeling excited. Shahaf wondered whether she could teach a computer to accurately assess how funny a caption might prove to be—and in the process, crack one of the most difficult challenges in machine learning. Sarcasm, wordplay, and other tools of humor have perplexed AI systems for decades. At Microsoft, teaching machines and software to get the joke is important for things like the Skype Translator, which is designed to let users speak to each other in different languages and hear translations on the fly.

For the study, Shahaf fed cartoons and captions from the New Yorker’s database into the system and trained it to find the funniest choices among captions that make similar jokes. She relied partly on crowdsourced input from contract workers, using Amazon.com’s Mechanical Turk. Then she moved to the harder task of ranking jokes. Because typical computer vision software is designed for photos, not drawings, the researchers had to manually describe what was pictured in each cartoon. They organized this into two categories: the context and the anomalies.
Read more... 

Source: Bloomberg