Natural language processing-based emotion and usage analysis of an AI-powered chatbot for epilepsy support

Area of Science:

• * Artificial Intelligence in Healthcare
• * Natural Language Processing (NLP)
• * Digital Mental Health Support

Background:

• * AI-powered chatbots are increasingly used for patient education and mental health support.
• * Their effectiveness in epilepsy care is underexplored, necessitating investigation into user interaction patterns and emotional impact.

Purpose of the Study:

• * To examine text-based interactions between users and an epilepsy-focused chatbot (EpiloBot).
• * To understand conversational patterns and emotional trends using NLP techniques.
• * To assess the chatbot's influence on user emotions and engagement.

Main Methods:

• * Analysis of text interactions from 23 participants (10 people with epilepsy, 10 caregivers, 3 healthcare professionals).
• * Utilized JMedRoBERTa for message classification into content clusters and DeBERTa for emotional shift tracking.
• * Employed UMAP clustering and k-means for query differentiation; sentiment analysis focused on frequent users (n=5).

Main Results:

• * Identified two primary content clusters: 'Medical Inquiry' (treatment-related) and 'Exploratory Use' (personal/emotional).
• * Sentiment analysis indicated proactive learning messages correlated with increased positive emotions.
• * Introspective, self-evaluative messages were associated with reduced positivity.

Conclusions:

• * NLP analysis of chatbot interactions provides insights into patient engagement and emotional support needs in epilepsy care.
• * Chatbots should balance medical guidance with psychological support for chronic disease management.
• * Future AI development should prioritize context-aware, empathic responses for optimized patient outcomes.