Predictive text errors inspire AI comedians

"YOU'RE my knight in shining armpit." "Do you want to come over and watch some mother?"

As anyone who has used them knows, text prediction systems in phones, search engines and word processors often get things wrong, sometimes hilariously. A team led by Alessandro Valitutti at the University of Helsinki in Finland want to harness that to liven up our interactions with software. The researchers reckon they can write programs that mimic this inadvertent humour.

Making computers funny is tricky, warns Mike Cook, a computer scientist at Imperial College London. "With research on humour, there is always that danger that by analysing it too closely it loses some magic."

But the Finnish team think autocorrect humour is ripe for automation, because the funniest mistakes are based on "simple and unintentional puns", says researcher Hannu Toivonen.

Their software, which will be presented at an AI symposium in Arlington, Virginia, next month, plays the humour card in two steps. The first is timing - if the software turned words in every sentence in an application into offbeat puns, it would just be tiresome. So it changes words only occasionally, interjecting at random intervals. Second, it derives comical terms by plundering dictionaries of related words on subjects that are favourite comedy staples, such as religion and sex.

"It might be funny or not, we cannot control it," admits Valitutti - but the idea is that sometimes it will be. One application for the software, he suggests, is to inject such humour into a reminder system, so instead of being told your cakes are about to burn, you might get an absurdly comical message that still encourages you to turn off the oven.

Cook thinks they might be onto something. "By understanding what makes these accidental slip-ups so funny, we open up avenues for our software to entertain us on a daily basis," he says. "That's exactly the kind of thing I'd love this research to unlock - more ways to make us laugh, and a new way for AI to improve our day-to-day life."

The team is not limiting their work to text. By combining it with a speech recognition system, they hope to enable a computer to anticipate and complete your spoken sentences with a witty remark.

If you would like to reuse any content from New Scientist, either in print or online, please contact the syndication department first for permission. New Scientist does not own rights to photos, but there are a variety of licensing options available for use of articles and graphics we own the copyright to.

Have your say

Only subscribers may leave comments on this article. Please log in.

Only personal subscribers may leave comments on this article

Subscribe now to comment.

All comments should respect the New Scientist House Rules. If you think a particular comment breaks these rules then please use the "Report" link in that comment to report it to us.

If you are having a technical problem posting a comment, please contact technical support.

Source: http://feeds.newscientist.com/c/749/f/10897/s/245a051b/l/0L0Snewscientist0N0Carticle0Cmg216288650B90A0A0Epredictive0Etext0Eerrors0Einspire0Eai0Ecomedians0Bhtml0Dcmpid0FRSS0QNSNS0Q20A120EGLOBAL0Qonline0Enews/story01.htm

apple dividend snow white and the huntsman snow white and the huntsman rupaul drag race walking dead comic kratom broncos