The COAST project aims to develop end-to-end AI and data systems for targeted surveillance and management of COVID-19 and future pandemics in Uganda. This is achieved through the collaborative effort of four workstreams focused on: (1) mining text and audio data to understand community perceptions; (2) developing detection and diagnostic tools for improved healthcare delivery; (3) modeling and forecasting disease transmission to inform policy; and (4) analyzing air quality and its relationship to COVID-19 risk.
we focused on four key objectives:

1. Collecting and transcribing radio data from different regions of Uganda.
2. Using the transcribed data to train and develop Automatic Speech Recognition (ASR) and Keyword Spotting models for three widely spoken Ugandan languages — Luganda, Acholi, and Lumasaaba.

3. Analyzing the radio data for COVID-19-related mentions, particularly around vaccination and public awareness.

4. Training the Luganda ASR model using transcribed radio data alongside Common Voice data to improve speech recognition accuracy.

Recorded radio broadcasts provide a rich source of information for capturing the voices of marginalized communities, especially women and those without internet access. Through qualitative and quantitative analysis, we gained insights into their lived experiences, concerns, views on pandemics like COVID-19 and Ebola, and their perspectives on future global health challenges.