Chinese Medical Dialogue: 792,099 Patient-Doctor Q&A Pairs Across Six Specialties

Comprising 792,099 Chinese-language question-answer pairs from medical consultations. It was created by erhwenkuo and converted from a GitHub repository, with data organized into six specialties including Internal Medicine (220,606 pairs) and Pediatrics (101,602 pairs).

Use Cases

• Fine-tune a large language model on instruction/input/output triads to simulate medical consultations in Chinese.
• Analyze the distribution of patient questions and doctor responses across six specialties like Andriatria and Surgical.
• Train a text classification model to route patient input text to the correct medical specialty.
• Build a sequence-to-sequence model for generating medical advice based on patient input descriptions of symptoms.

Strengths

• Contains 792,099 total dialogue pairs, providing substantial text data.
• Data is categorized into six distinct medical specialties with specific counts, such as 220,606 pairs for Internal Medicine.
• Sample data structure shows specific fields like 'instruction', 'input', and 'output' for training conversational AI.

Limitations

• The original data source is a GitHub conversion, which may introduce formatting inconsistencies or lack provenance details.
• No column definitions, sample data, or file format details are provided in the input, limiting initial assessment.
• Potential for label or specialty misclassification exists without validation of the original data collection method.

Provenance

Source

Converted from Toyhom/Chinese-medical-dialogue-data GitHub repository.

Collection Method

Conversion from an existing repository; original collection method unknown.

Freshness

Last updated on 2023-09-27.