Practical AI NLP to help pregnant mothers in Kenya
Aug 17, 2021
Jay Patel and Sathy Rajasekharan from Jacaranda Health share their mission to revolutionize maternal care in Kenya using AI-powered NLP solutions. They discuss the alarming maternal mortality rates and how SMS technology helps prioritize urgent messages from pregnant mothers. The duo explores the challenges of data labeling and collaboration in developing effective communication platforms. They also highlight the importance of navigating Kenya's linguistic diversity to enhance health responses, ultimately aiming to ensure safe childbirth and newborn care.
AI Snips
Chapters
Transcript
Episode notes
Accidental Two-Way Communication
- Jacaranda Health accidentally discovered a latent demand for pregnancy information via SMS when mothers started asking questions.
- This led them to develop NLP tools to triage and prioritize these incoming messages, as some indicated serious danger signs.
Chatbot Failure
- Initial attempts to automate conversations with chatbots failed, as mothers didn't like the impersonal responses.
- This highlighted the need for human interaction and careful consideration of language complexity in healthcare communication.
NLP Model Selection and Refinement
- Start with existing commercial NLP models before building custom solutions, testing with labeled data to find the best fit.
- Refine intent classification by finding the right level of granularity, not too broad or too specific, through iterative testing.