Problem
Research questions and friction points this paper is trying to address.
conversational diagnosis prediction
multi-person dialogue
emergency medical service
medical dialogue corpus
streaming clinical conversations
EMSDialog: Synthetic Multi-person Emergency Medical Service Dialogue Generation from Electronic Patient Care Reports via Multi-LLM Agents
🤖 AI Summary
This study addresses the scarcity of multi-party clinical dialogue data in emergency medical services (EMS), as existing datasets are largely confined to dyadic interactions and lack structured diagnostic annotations aligned with clinical workflows, thereby hindering research on dynamic diagnosis prediction. To bridge this gap, the authors propose the first multi-agent synthetic dialogue generation framework tailored for EMS scenarios, integrating topic-flow planning with collaborative large language model (LLM) agents. The framework employs rule-based fact consistency verification and an iterative self-refinement mechanism to produce structurally coherent, high-quality dialogues. The resulting EMSDialog dataset comprises 4,414 conversations covering 43 distinct diagnoses, multiple participant roles, and turn-level topical structures, significantly enhancing the accuracy, timeliness, and robustness of diagnostic prediction models while filling a critical void in multi-participant clinical dialogue resources.
📝 Abstract
Conversational diagnosis prediction requires models to track evolving evidence in streaming clinical conversations and decide when to commit to a diagnosis. Existing medical dialogue corpora are largely dyadic or lack the multi-party workflow and annotations needed for this setting. We introduce an ePCR-grounded, topic-flow-based multi-agent generation pipeline that iteratively plans, generates, and self-refines dialogues with rule-based factual and topic flow checks. The pipeline yields EMSDialog, a dataset of 4,414 synthetic multi-speaker EMS conversations based on a real-world ePCR dataset, annotated with 43 diagnoses, speaker roles, and turn-level topics. Human and LLM evaluations confirm high quality and realism of EMSDialog using both utterance- and conversation-level metrics. Results show that EMSDialog-augmented training improves accuracy, timeliness, and stability of EMS conversational diagnosis prediction.
Innovation
Methods, ideas, or system contributions that make the work stand out.
multi-agent dialogue generation
ePCR-grounded synthesis
topic-flow planning
conversational diagnosis prediction
synthetic EMS dataset
🔎 Similar Papers
No similar papers found.
