Maternal Health Question/Answer Dataset for Training AI models in English, Swahili, Luganda and Runyankore

The DoD_chat dataset is a curated parallel corpus designed to support research and development of multilingual healthcare conversational systems, machine translation models, and Large Language Models (LLMs) for low-resource settings. The dataset contains aligned question-and-answer pairs in four languages, covering maternal health topics such as prenatal care, childbirth, postpartum care, nutrition, danger signs, and newborn health.

A key contribution of the dataset is its preparation for the analysis of how health information is communicated across different languages within a developing-world context. By maintaining semantic equivalence across all language versions, the corpus enables comparative studies of information presentation, linguistic variation, cultural adaptation, and healthcare communication strategies.

To facilitate the development of modern LLMs and conversational AI systems, the dataset was carefully engineered to minimize the use of first-person and collective pronouns such as *I*, *me*, *we*, and *us*. This design choice promotes a neutral, instructional, and professional communication style that is more suitable for model training and deployment in healthcare advisory systems.

Furthermore, all responses were translated and adapted within the context of a medical professional providing advice through a chatbot or Large Language Model. Rather than performing literal translations, the dataset emphasizes context-preserving medical guidance, ensuring that responses remain clinically appropriate, culturally relevant, and conversational in nature across all languages.

The resulting corpus provides a high-quality multilingual resource for research in machine translation, multilingual NLP, health communication, cross-lingual semantic analysis, and the development of healthcare-focused LLMs capable of delivering consistent and reliable maternal health information across diverse linguistic communities.