Korean Medical Dialogue Dataset from MedText and ChatDoctor

Ko Medical Chat is a Korean-language medical conversation dataset created by squarelike. It was constructed by converting the English MedText and ChatDoctor datasets into Korean. The dataset was last updated on Hugging Face in August 2023.

Use Cases

• Fine-tuning large language models for Korean medical question-answering using doctor-patient dialogue pairs.
• Training a classifier to identify conversation roles (e.g., patient, doctor) within the transcribed clinical text.
• Developing intent recognition models for Korean medical consultations based on user query patterns in the dialogues.
• Benchmarking machine translation quality for medical domain text between English source datasets and this Korean conversion.

Strengths

• Derived from established medical text sources (MedText and ChatDoctor), providing a foundation in clinical content.
• Specifically addresses the need for Korean-language resources in the medical AI domain.

Limitations

• No information provided on dataset size, row count, or sample quality metrics.
• Potential for translation artifacts or loss of nuance from converting English medical datasets to Korean.
• Lack of column details prevents assessment of structure, features, or potential biases.

Provenance

Source

Converted from the MedText and ChatDoctor datasets.

Collection Method

Conversion of existing English medical conversation datasets into Korean.

Freshness

Last platform update was 2023-08-19.

Geography

Language coverage is Korean; geographic origin of source data is unspecified.