Doctor-patient Conversational Dataset

Home » Case Study » Doctor-patient Conversational Dataset

Project Overview:

Objective

The Doctor-patient Conversational Dataset aims to create an extensive, annotated audio dataset that accurately represents a wide range of medical consultations. The primary objective is to develop this dataset so that it will be instrumental in training AI systems to understand and process healthcare-specific dialogue, thereby enhancing patient care and support.

Scope

The project encompasses various medical specialties, ranging from general practice to more specialized fields like cardiology and neurology. It includes diverse patient demographics to ensure a comprehensive representation of real-world medical conversations.

Sources

Data is sourced from consenting participants in various healthcare settings, ensuring confidentiality and ethical compliance. Collaborations with hospitals and clinics across different regions provided access to a rich pool of audio conversations.

Data Collection Metrics

Total Audio Hours Collected: 1,500 hours
Number of Unique Conversations: 10,000
Participant Demographics: 45% male, 55% female, ages ranging from 18 to 85
Medical Specialties Covered: 15, including General Practice, Pediatrics, Oncology

Annotation Process

Stages

Transcription: Converting audio files to text.
Categorization: Classifying conversations based on medical specialty and topics.
Entity Tagging: Identifying and tagging medical terms, symptoms, and medications.

Annotation Metrics

Total Conversations Annotated: 10,000
Total Annotations: 500,000
Average Annotations per Conversation: 50

Quality Assurance

QA Metrics

Accuracy of Transcription: 98%
Consistency in Categorization: 95%
Precision in Entity Tagging: 97%

Conclusion

The Doctor-Patient Conversational Dataset project is a landmark initiative in the intersection of healthcare and AI. By providing a rich, well-annotated dataset, it paves the way for advancements in conversational AI, ultimately leading to more effective and empathetic patient care.