Key Information Extraction Dataset for Doctor-Patient Conversations in Consultation Rooms

In the medical field, doctor-patient conversations contain a wealth of key information that is crucial for improving diagnostic efficiency and patient satisfaction. However, current technological solutions often lack accuracy and reliability when identifying and extracting this key information, especially in noisy environments and when dealing with multilingual and medical terminologies. This dataset aims to improve key information extraction capabilities in audio dialogues, meeting the specific needs of the healthcare industry. Data collection was conducted using professional recording equipment in consultation room environments to ensure realistic scenario simulation. Quality control involved a three-round annotation process, consistency checks, and medical expert review, with the annotation team consisting of professionals with medical backgrounds. Data preprocessing includes audio denoising, segmentation, and transcription, with the final data stored in WAV format and organized in JSON format as transcribed dialogue texts.