A Deep Learning-Based Chatbot to Enhance Maternal Health Education
John Batani, Elliot Mbunge, Lipuo Leokana
Abstract
Maternal mortality remains a global concern, with resource-constrained countries disproportionately affected due to inherent challenges in such countries, like underfunding, distant health facilities, lack of access to maternal health education and inequitable access to maternal health services. Though medical chatbots are gaining popularity, resource-constrained countries lag, and there is a dearth of chatbots specific to maternal health education in local languages. Therefore, this study utilised natural language processing to develop a maternal health education chatbot using a feedforward deep neural network. The model was trained using three local African languages (Sesotho, Shona and Ndebele) and English, and the chatbot was deployed using the Flask server through a web app to present a friendly interface to users. The training and evaluation losses reached zero, while the training and evaluation accuracies reached 100%.