Crowdsourcing improves smartphone texting

Smartphones may soon be a lot better at predictive texting -- finishing sentences for you -- with help from crowdsourcing, British and U.S. researchers say. Predictive text software in current smartphones often has a hard time with texts and voice commands when a user attempts words or phrases not already in the phone's database, they said. To see if they could improve that, Keith Vertanen of Montana Tech in Butte and Per Ola Kristensson at the University of St. Andrews in Britain turned for help to workers of the Amazon Mechanical Turk, a crowdsourcing Internet marketplace where computer programmers can utilize human intelligence to facilitate tasks computers aren't up to yet. The researchers' goal was to improve a predictive system used in Augmented and Alternative Communication devices which help disabled people communicate by painstakingly typing out words interpreted from their muscle twitches or blinks, NewScientist.com reported. Mechanical Turk workers were paid to imagine phrases they might need if they had motor neuron disease or cerebral palsy, and they produced nearly 6,000 useful phrases. By incorporating the phrases in a text database, the result was a system requiring 11 percent fewer keystrokes than a standard AAC device, the researchers said. That's a big improvement for anyone struggling over every word, Kristensson said, noting the process should also work when applied to more standard text and speech recognition systems.