Key points are not available for this paper at this time.
Chatbot-based systems are recognised in literature as an effective tool to support chronic diseases self-management. However, the core of most of this work is on the description of the application domain and on the motivations behind the adoption of recommendation systems that exploit chatbot to mediate the interaction with users, but they fail in providing sufficient details on the system architecture and on the technology adopted. Moreover, they are usually designed with a strong focus on the specific pathology, and a reference architectural solution that can be adopted in different contexts is missing, thus making the work useful only in the domain it is devised for. In this paper we provide a framework for developing recommendation systems based on chatbots that is meant to be applied in different scenarios. The framework is composed by a back-end recommendation engine that autonomously computes the user's adherence profile to prescription, and proactively provides motivational feedback to the user through the application front-end based on a chatbot. The chatbot is also meant to collect and aggregate data for profiling the individual health and habits. To demonstrate the feasibility of our framework, we present a recommendation system, based on a Telegram chatbot, that has been developed and trained for managing hypertensive patients.
Montagna et al. (Thu,) studied this question.