Conversational interface for improved medication adherence

A project led by Dr Carmelo Velardo at the University of Oxford.

New technologies are increasingly delivering monitoring and support for the effective management of long-term conditions and chronic diseases. New clinical pathways facilitated by such technologies are key to addressing the associated problems of frailty and multiple co-morbid conditions, and the challenges of delivering health services to the rising number of people with these conditions.

Previous work has shown that using brief SMS messages can support brief behaviour change with targeted educational messages. Whilst short-term practicalities for wide reach of messaging requires SMS messaging, modern communication technologies are now available that take advantage of the pervasiveness of mobile platforms and the capabilities of smart mobile devices.

These technologies are designed for switching messages to multiple delivery platforms and that allow innovative methods of communication, including the possibility of moving to Conversational interfaces (CI), i.e. systems that provide an extension of human-computer interactions as a conversation.

Several CI have recently been developed for symptom checking with the aim of replacing or pre-screening medical consultations. However, many lack proper validation studies.

This project entails the collection and analysis of natural language data extracted from  un-directed interaction of patients with diabetes with an automatic system. The main goal of the study is to analyse the data collected in order to derive a functional, autonomous, conversational agent that is able of interacting and responding via free text to questions about diabetes. To that aim, we will collect data from patients with diabetes using retrospective analysis of self-help, online forums, and by recording the interactions of real patients with a simulation of conversational agent performed by a trained researcher (Wizard of Oz experiment). The set of information collected will help defining the topics of interest to patients with diabetes, and will provide the content that will be used to train the conversational agent.

<< Back