Speaker Role Identification in Clinical Conversations.

Area of Science:

• Medical Informatics
• Computational Linguistics
• Artificial Intelligence

Background:

• Patient-clinician communication is vital for healthcare outcomes.
• Manual analysis of clinical interactions is time-consuming and challenging.
• Existing automatic speech recognition systems lack speaker role identification capabilities.

Purpose of the Study:

• To investigate the effectiveness of Large Language Models (LLMs) for Speaker Role Identification (SRI) in clinical settings.
• To compare LLM performance with and without accurate diarization identifiers.
• To assess the impact of identifier corruption on SRI performance.

Main Methods:

• Utilized BERT, a Large Language Model, for Speaker Role Identification (SRI).
• Employed veridical turn segmentation and diarization identifiers.
• Fine-tuned models with varying levels of identifier corruption to evaluate performance.

Main Results:

• BERT achieved 82% accuracy and 82% F1-score using linguistic signals alone.
• Incorporating accurate diarization identifiers improved performance to 95% accuracy and 95% F1-score.
• LLM performance was robust even with corrupted diarization identifiers.

Conclusions:

• Fine-tuned LLMs demonstrate significant potential for accurate Speaker Role Identification (SRI) in clinical transcripts.
• LLMs offer an automated solution for analyzing patient-clinician communication dynamics.
• This approach can enhance research in clinical discourse and healthcare outcomes.