Psychotherapist remarks' ML classifier: insights from LLM and topic modeling application.

Area of Science:

• Psychology
• Computer Science
• Linguistics

Background:

• Systematic analysis of therapist communication is limited.
• Understanding recurring language patterns can improve clinical practice, supervision, and training.
• Machine learning (ML) offers new avenues for analyzing complex therapeutic dialogues.

Purpose of the Study:

• To develop an ML classification model for analyzing topics in therapist remarks.
• To explore the intersection of ML and psychotherapy through topic modeling.
• To compare thematic structures in classical and modern psychotherapeutic approaches.

Main Methods:

• Applied BERTopic, an ML-based topic modeling technique, to therapist dialogues.
• Utilized dimensionality reduction, clustering, and expert refinement for topic analysis.
• Trained an ML classifier using identified topics and tested on a case study.

Main Results:

• Identified common and stable topics across classical and modern therapists.
• Revealed distinct thematic differences between therapist groups.
• Demonstrated robust classifier performance in distinguishing therapeutic discourse patterns.

Conclusions:

• Automated topic modeling with expert input effectively uncovers therapist language patterns.
• The publicly available model has broad applications in psychotherapy research and training.
• Topic modeling is a valuable tool for advancing ML in psychotherapy and understanding therapist communication.