A few weeks ago, I met Robert Munro, a computational linguist at Stanford University. Rob seems to have done many things so far in his still-young career (see his list of publications), but most recently of note is Mission 4636, an SMS-based crowd-sourcing translation tool intended for disaster relief. I caught a clip on NPR’s Morning Edition the other day featuring this very technology – the idea of using crowds for translation rather than computer algorithms. We’ve all used Yahoo!’s Babel Fish or Google Translate and been dissatisfied either with their language selection, the translation, or both. These web tools are decent for translating single words or even whole web-pages and long passages when the output is not required to be eloquent. But, they have their limitations, and in the wake of the January 12th Haitian earthquake, Rob and his colleagues identified a real need for something better. The basic problem is that cell phone networks in Haiti became overwhelmed in the aftermath. When this happened, little data other than SMS messages could pass through (a problem also identified by I’MOK, who we covered a few weeks ago). Thousands of survivors were sending messages for help, but who was sent there to receive them and manage the recovery process? The non-Creole speaking U.S. Military.
Within hours of the earthquake, 4636 was established as a short code to which one could text messages for help. Within 2 days, Mission 4636 was up and running, capturing those messages, distributing them to thousands of Creole-speakers worldwide, and sending back their translations to aid workers on the ground in Haiti. The turnaround time was about 10 minutes, and the system helped deliver the first food and aid to over 10,000 earthquake survivors.