Related Experiment Video
Updated: Jun 23, 2026
Virtual Agent for Real-Time Motivational Interviewing by Integrating Adaptive Nonverbal Behavior and Language Models
Published on: December 23, 2025
Automating the roter interaction analysis system for medication counseling: A transformer-based deep learning
Ayako Mori1, Satoshi Watabe2, Izumi Kato3
1Education Research Center for Clinical Pharmacy, Faculty of Pharmaceutical Sciences, Hokkaido University, Kita 12, Nishi 6, Kita-ku, Sapporo, 060-0812, Japan; Laboratory of Clinical Pharmaceutics & Therapeutics, Division of Pharmasciences, Faculty of Pharmaceutical Sciences, Hokkaido University, Kita 12, Nishi 6, Kita-ku, Sapporo, 060-0812, Japan.
Automated Roter Interaction Analysis System (RIAS) classification for medication counseling dialogues is feasible using Japanese transformer models and AI data augmentation. While effective for common expressions, nuanced affective categories require multimodal approaches for improved accuracy.
Area of Science:
- Natural Language Processing
- Medical Communication Analysis
- Artificial Intelligence in Healthcare
Background:
- The Roter Interaction Analysis System (RIAS) is the standard for medical communication analysis.
- Automated RIAS coding is challenging due to imbalanced code distributions.
Purpose of the Study:
- To develop an automated RIAS classification system for medication counseling dialogues.
- To evaluate AI-based data augmentation for mitigating class imbalance in RIAS coding.
Main Methods:
- Fine-tuning five transformer models for 44-class RIAS classification.
- Employing AI-based data augmentation to improve model generalizability.
- Evaluating model performance on both AI-augmented and real-data-only test sets using accuracy, macro F1, and weighted F1 metrics.
Main Results:
- ELECTRA achieved the highest performance metrics (accuracy: 0.7875, macro F1: 0.6561, weighted F1: 0.7835) on the real-data-only test set.
- Models performed better on AI-augmented data, particularly for minority classes.
- Challenging categories included semantically ambiguous, context-dependent, and affective expressions.
Conclusions:
- AI-based augmentation can successfully mitigate class imbalance in automated RIAS classification.
- Text-only approaches struggle with nuanced affective categories, suggesting a need for multimodal analysis.
- Automated RIAS analysis shows promise for pharmacy education and patient-centered care.