INTRODUCTION: Behavioural support has been shown to possess high efficacy in aiding smokers to stop smoking at a level at least similar to nicotine replacement therapies (NRT) while causing no adverse physiological effects. Furthermore, the effects of behavioural interventions and NRT appear to be roughly additive. The effectiveness of interventions appears to increase with the frequency of contact with an eHealth application, with daily contact showing higher efficacy compared to weekly sessions with a trained stop-smoking specialist. Previous automated counsellors based on a motivational interview approach, lapse preparation, and lapse management are rigid in their therapeutic path, with limited ability to reflect the needs and the specific situation of a given patient. AIMS: This report aims to describe a possible approach to developing a more engaging, patient-tailored automated counsellor based on recent advances in the natural language processing (NLP) field that should make remote chat-based counselling easier for professionals while gathering data for the NLP model, which should ultimately be able to conduct the therapy on its own. METHODS: The core of the approach lies in utilizing the Text-to-text transfer transformer (T5). T5 is, in essence, a set of neural network models aimed at tasks formulated as an expected textual response to a given textual input. These models can be utilized to – at first – suggest answers to counsellors in live chat sessionswith patients. Actual answers from these sessions would subsequently be used to fine-tune the models and ultimately provide high-quality counselling without human intervention on the therapist’s side. CONCLUSION: The article presents a novel approach to internet-delivered smoking cessation cognitive-behavioural therapy utilizing a powerful artificial neural network NLP model acting as a conversational agent and a data collection protocol with usage incentives for both smoking cessation experts and smokers.

Czech Technical University Faculty of Electrical Engineering Department of Cybernetics Prague Czech Republic

References provided by Crossref.org

Literatura