Signs of progress

WHERE the internet has yet to take firm root, people plump for the next-best thing. In many emerging markets this means text messaging. Customers have embraced the short message service (SMS) not just to communicate with each other, but also to get weather forecasts, bus schedules and traffic information, or to vote in television talent shows. Authorities use it for public announcements. Companies text targeted advertising. India even has a rudimentary SMS-based social network.

The reason texting isn’t even more popular in such places has to do with the fact that—despite indignant claims in some quarters that texting is killing literacy—texters must be able to read and write. One could get a phone’s text-to-speech program to read incoming text messages out loud, a feature available even on simple devices. But that still leaves out the hard of hearing. In Tunisia, for instance, perhaps 90% of the 200,000 or so deaf are also illiterate, estimates Mohamed Jemni, a computer scientist at the Ecole Supérieure des Sciences et Techniques in Tunis.

So Dr Jemni’s team built a system which uses natural-language processing to translate text messages into sign-language animations delivered in the form of a multimedia message (MMS). MMSs work on most phones and the actual data-crunching is performed not on the device but on the EUMEDGrid, a distributed-computing facility.

Dr Jemni claims that around 1,000 people already use his program, MMSSign, on Tunisia Télécom, the country’s national operator. Before the system can be adopted in other countries facing a similar illiterate-deaf predicament, though, it has to learn hundreds of signing languages and dialects in use worldwide. In order to make that possible, the team is designing a web interface to make building and updating the text-to-signing dictionaries easier. (Signing-to-text conversion is also in the works, though further off.)

As for personal communication, video streaming would help, but it demands fast mobile-broadband networks and expensive phones. And whereas a choppy or grainy image might be acceptable in a video call optimised for voice, research suggests that signing is hard to read on small screens using existing video-compression algorithms. So another research team at the University of the Western Cape in South Africa, led by William Tucker, is busy figuring out how to optimise mobile communications for the deaf. They are examining how to adapt compression algorithms for video streaming as broadband connections grow more accessible. In the meantime, they too are compiling digital dictionaries on grid computers, with a view to making two-way translation between the written word and sign language possible.

Before long, then, hitherto ill-served customers may at last begin to delight in the joys of texting. They may rejoice less at the inevitable sign-language spam.

First published by The Economist’s Babbage blog: [html] [pdf]