JMedBench-Train / README.md
stardust-eques's picture
Create README.md
2d538f1 verified

Introduction

This is a modified dataset from a part of JMedBench. It consists of 674,954 lines of Japanese and English medical text. The column is named "text".

Only the train subset is extracted from the data source and then concatenated, where the question-answering samples are made into one sentence by

Questions:{Question}Answer:{Answer}

for English samples and

質問:{Question}回答:{Answer}

for Japanese samples.

No other modification was applied.

Limitations

Please read the original data sources.

Licenses

Please follow the original data sources.

Intention

This dataset is meant to be used for the continual pretraining of medical LLMs.

No other usage is recommended.