Researchers from the University of Toronto and Microsoft are exploring ways to use technology – specifically, mobile phones – to crowdsource the job of digitizing hand written documents word by word. Phones such as low-cost Samsungs and Nokias have the ability to display photos as bitmapped text messages. A business model that crowdsources digitization of these documents would enable phone users to receive the text, and type and send the word back, creating an opportunity to earn supplemental income as well as contributing to a much larger task.
Low-tech phone technology allows images to be sent as bitmaps in text messages, opening up applications for the world’s poorest.
When it comes to mobile communications, there’s still a lot of room for innovation at the bottom. In Bangalore, India, researchers from the University of Toronto and Microsoft are now imagining new business models for the world’s poorest phone owners by adapting a little-known protocol that can receive pictures as bitmapped text messages. The technology could readily be used in the roughly 1.5 billion low-end Nokia and Samsung phones in circulation.
The researchers have shown how to use the technology to crowdsource the task of digitizing handwritten documents word by word, a type of work that anyone with an inexpensive phone could do for extra money. A handwritten word is displayed as the image—resembling retro graphics from 1980s Atari games—and the phone owner types in the word, contributing piecemeal to a larger job.
“Crowdsourcing on phones really has potential to provide substantial income for people who are very poor and have a lot of time on their hands,” says Ed Cutrell, a computer scientist at Microsoft Research India in Bangalore. “That’s one of the main things we’re interested in here: are there means to provide supplemental income to people who don’t have computer and Internet access?”
Tools such as Amazon’s Mechanical Turk already allow such applications and can be accessed using smart phones, but they generally require a full-fledged computer and Internet connection. In fact, though about one-third of Mechanical Turk workers are based in India, studies have found that these workers are generally college educated, make double the national average income, and are using computers, not mobile phones.
By contrast, the protocols involved in the new research, which are part of Nokia’s Smart Messaging Service, allow binary picture messages with a resolution of only 74 by 28 pixels. This would allow a rough image to be displayed on basic $20 phones with 1.3-inch black-and-white screens—and over SMS, which is a widely available protocol. “It’s an old technology that Nokia had implemented,” says Aakar Gupta, lead author of the paper, which is to be released at a conference today. “Most people had just forgotten about it.”
Indeed, Nathan Eagle, a Harvard professor and CEO of Jana (formerly known as Txteagle), which uses low-end phones as a platform for applications like marketing and surveys in poor regions of the world, says he’d never heard of the image protocol for SMS messages before the paper. “When people think about smart phones, they think about devices like the iPhone and applications in the App Store or Android marketplace,” he says. “But enabling low-end phones to display bitmap images, on a billion-plus devices—that’s exciting.”
Eagle can see a number of possible applications for the technology. “It’s a cool idea, and frankly, it’s worth broadcasting that this protocol exists,” he says. “If you are able to send a bitmapped image to low-end handsets, there are a lot of things you can do.”
Other applications could include sending bar-code identifiers as a banking security measure, say the Microsoft and Toronto researchers. But they are focusing initially on the crowdsourcing idea, which they call mClerk. In a five-week pilot test, the researchers had phone owners digitize handwritten words in the local language, Kannada, in a region four hours from Bangalore.
Within five weeks, 239 users had completed 64,000 tasks, for a total of 25,000 digitized words in a handwritten document that had been chopped into thousands of images of individual words.
The researchers paid participants with phone minutes, not cash, and estimate that this form of payment for a person working just two hours a day every day could equate to about $21 a month—12 percent of the average monthly wage in the region. Gupta says the concept could extend to all sorts of handwritten things, like information on medical forms—perhaps allowing for the distributed digitization of medical records.
The researchers have no immediate plans for commercialization or licensing the technology, although they will develop it further over the summer.
Eagle cautions that crowdsourced transcription may not end up being the killer app for displaying images on text messages. But he says one could envision developers writing simple games or educational tasks, or even sharing simple images as part of text-message-based social networks. “It’s great to see more people starting to think about other uses for these low-end handsets beyond standard phone calls and text messages,” he says.