This is a virtual coach system that coaches users into being more physically active and stop smoking.
PerfectFit is a text-based virtual coach system that helps the users in quitting smoking and being physically more active. The software provides the backbone of the system and the logic for intervention administration.
The core of the system is the open-source text- and voice-based contextual assistant [Rasa] https://github.com/RasaHQ/rasa). Rasa is a machine learning framework, that provides automated conversations based on personalized contexts, that can be connected to messaging and voice services through standard interfaces. Perfect Fit leverages the Rasa framework, for the development of custom conversation flows. Such logics are used to guide the VC behaviour.
An SQL database has been developed and connected to the system to store and use the information needed to tailor the intervention to the user's needs. . To allow the VC to asynchronously initiate the conversation, a scheduler based on Celery system is available for tasks queuing. The scheduler can be used to ask questions for collecting information in predefined timepoints used to check and track the status of the user. In the configuration described, Perfect Fit is designed as an intelligent conversational agent for smoking cessation, that can be integrated with different messaging channels using standard or custom interfaces. Perfect Fit can be used to conduct research on VCs adoption or it can be used as an add on to existing widespread messaging services for the provision of virtual coaching.
The conversational engine is connected to the NiceDay App, an application that provides the possibility to chat with a therapist to be supported in mental health care. In this configuration, the therapist is substituted by the VC, with which the users can interact through the chat. The two systems have been integrated using connectors developed ad hoc.
Targeting key risk factors for cardiovascular disease in at-risk individuals using a personalized and adaptive approach