Making Ugandan Community Radio Machine-readable Using Speech Recognition Technology
PARTNER: UN Agencies, Stellenbosch University
PROGRAMME AREA: Real-time evaluation
LAB: Pulse Lab Kampala
PROJECT MICROSITE: http://radio.unglobalpulse.net/uganda/
DOWNLOAD SUMMARY: UN Global Pulse, “Public Radio Content Analysis Tool”, Big Data Tools Series no. 1, 2016 [PDF]
Radio is a vibrant medium for public discussion in Uganda. Talk shows and phone-ins hosted by community radio stations are popular ways for Ugandans to voice their needs, concerns and opinions—particularly in rural areas. Analysis of radio content therefore presents an opportunity to take the pulse of populations excluded by the digital divide. In 2014 there were 216 registered FM radio stations across the country, broadcasting on 299 different transmitters (source: Uganda Communications Commission). Pulse Lab Kampala worked on a toolkit that makes public radio broadcasts machine-readable through the use of speech recognition technology and translation tools that transform radio content into text.
This project seeks to support the Government of Uganda and development partners in incorporating the voices of Ugandan citizens into the development process. Better understanding of public opinion will support bridging the gap between policy and implementation of development programmes. The project has been initiated by Pulse Lab Kampala in partnership with Makerere and Stellenbosch (South Africa) Universities, with the support of the Embassy of Sweden in Uganda, and seeks to expand the scope of sentiment analysis in Uganda to rural areas developing a toolkit to track public conversations from radio. In Phase I of the project, recordings of public radio content were converted into text automatically from Kampala (English and Luganda) and Gulu areas (Acholi and English).
In Phase II of the project, Pulse Lab Kampala in collaboration with UN agencies, NGOs and the Government of Uganda, is conducting a series of pilot case studies to assess the potential of the tool to inform on issues of relevance for the SDGs. This will be done by expanding the geographical coverage of broadcast analysis to new locations. A fully automated system will be developed to cut the time of the data flow from the radio broadcast to the dashboard to 1 day (from the time it is recorded to the time it appears in the dashboard developed as part of the tool).
Aditionally, protocols will be developed to ensure: (1) integration with other tools developed by Pulse Lab Kampala, (2) implementation of technical measures to address the risks associated with the development of the tool and (3) mechanisms that protect the privacy of individuals are in place. The tool is being used to identify trends amongst larger groups and communities with no intention of identifying opinions of individuals.