MedDialSpeech: Interaction-Aware Multi-Speaker Medical Dialogue Generation A fully-synthetic English doctor–patient dialogue audio benchmark for conversational ASR and speaker diarization under realistic interaction conditions: backchannels, overlap speech, fillers, and adversarial interruptions.

	Splits




	

Split

Profile

polite 195 Moderate clarification + backchannels (mode=polite).

adversarial 196 High-density interruptions / talk-over… See the full description on the dataset page: https://huggingface.co/datasets/Tomatohust/meddialspeech.