The United Nations often uses social media in many countries to get a sense of what people are in need for in terms of assistance or other concerns. However, many rural areas in places like Uganda don’t have as much access to the internet, so radio is still a popular source of information and voicing concerns. In order to address this, the U.N. is developing a radio-listening technology that can perform voice recognition from radio programs for specific terms, like “flooding.”
The problem, specifically for Uganda, was the language. Installing a listening platform only took a Raspberry Pi device, but the U.N. has also had to teach the computers to recognize Uganda-accented English, Acholi and Luganda, none of which were served by existing speech-recognition programs.
The current system is able to recognize up to 50 percent of words in a transcript, which the U.N. says is enough for it to flag a potential incident. But now they are looking to see if the radio analysis is helpful. How the tool might ultimately be deployed and who would have access to it won’t be decided until further data is gathered.
The ultimate goal it says is “to involve more voices of rural citizens in decision-making about where to send aid or how to improve services.”