2017 ICT4D Conference

Africa’s Voices Foundation (AVF) is a not-for-profit organisation, developed from cutting-edge research at the University of Cambridge. Our methods were developed to overcome the challenge of engaging with hard-to-reach citizens through leveraging existing communication channels such as radio and social media. In these channels, we curate conversations in order to create a context where the audiences feel a familiarity and a willingness to participate. We then use methods from data science, social science and our contextual knowledge to analyse large volumes of conversational data, combining insights derived from these conversations with socio-demographic data. This approach has allowed us to move beyond the ""what"" and ""who"" of surveys to also provide insight into ""why"" something matters – expressed in citizens’ own language and emic perspectives. We have so far worked with UNICEF Somalia, Oxfam in Kenya, BBC Media Action, Cambridge and Makerere Universities, and Emmy-award winning the Well Told Story, and Livity Africa. Topics have included health in Somalia and oil and gas extraction in North Kenya.

This presentation talks about the challenges and opportunities involved in analysing large volumes of conversational data in low resource languages in Africa.  There are currently few resources that aid the computational analysis of such large textual datasets and this makes the development of these resources both time- and resource-intensive. Some of the solutions we have developed so far to address this challenge includes building customised language banks, testing machine learning approaches for labelling data, and developing new tools to speed up the process of labelling data. Focusing on our ongoing work with UNICEF in Somalia and with Well Told Story in Kenya, the presentation outlines what are some of the lessons learned from this process – the challenges involved but also the opportunities of sharing resources among the ICT4D community to address thes knowledge gaps.