Systematic Review of Chatbot Development in Health and Education: Technologies, Frameworks, and Evaluation Metrics

Chatbots have emerged as valuable tools in public service domains such as healthcare and education, supporting information delivery, training, and user engagement. Despite growing adoption, development practices remain fragmented. This paper presents a Systematic Literature Review (SLR) of 27 studies to identify trends, challenges, and research gaps in chatbot development, focusing on technological approaches, frameworks, evaluation methodologies, and language distribution. Findings reveal a strong preference for open-source frameworks, particularly Rasa, valued for flexibility and scalability, while TensorFlow is often used for custom ML-based implementations. English dominates as the primary development language, with limited representation of Spanish, despite its global importance. Most studies apply quantitative (performance-based) metrics such as F1-score and accuracy; only one combines quantitative and qualitative evaluations, reflecting a lack of standardized assessment approaches. Additional gaps include the absence of hybrid evaluation frameworks, limited attention to Responsible AI principles, and scarce discussion on deployment requirements. Future research should prioritize multilingual chatbot models, integrate hybrid evaluation frameworks, and promote transparency to ensure scalability, ethical AI practices, and user-centered design.