Multilingual Conversational AI Framework for Empowering Maternal and Child Healthcare in Zimbabwe

The purpose of this study was to examine the potential benefits of a multilingual conversational AI framework for mother and child healthcare in Zimbabwe. This country has language hurdles towards healthcare access. By creating an artificial intelligence (AI) system that can understand Shona, Ndebele, and English, the research hopes to reduce communication gaps for linguistically varied communities. The study employed a mixed-methods approach, integrating both quantitative analysis of AI-patient dialogues and qualitative interviews with healthcare experts. ZimHealthBERT (a novel modification of BERT architecture), an AI model, was developed with transformer-based architectures, and a medical knowledge graph to ensure that responses were correct and suitable for a given culture. It also included cultural adaption layers. This model was trained online using the NVIDIA A100 GPU on RunPod Console. According to key results, prenatal care attendance increased by 21.9%, medication adherence increased by 20.6%, and patient comprehension improved by 35.5%. Healthcare workers reported 68.7% less time devoted to interpreting languages and a 30.9% increase in efficiency. In comparison to human interpreters, the research demonstrates greater management of culturally sensitive topics by providing a multilingual AI system specifically designed for African healthcare facilities. This represents cutting-edge contributions. The results demonstrate this framework's capacity to lessen healthcare disparities and imply that it can be expanded to additional multilingual developing countries. Future research will examine how to improve the system's accessibility in remote places through offline functionality, more cultural adaption enhancements, and possible integration with telemedicine platforms.