# MedText

### Paper

Title: `Towards Automatic Generation of Shareable Synthetic Clinical Notes Using Neural Language Models`

Abstract: [https://arxiv.org/abs/1905.07002](https://arxiv.org/abs/1905.07002)

MedText is a medical diagnosis dataset containing over 1000 top notch textbook
quality patient presentations and diagnosis/treatments. The 100 most common diseases
and the 30 most common injuries people go to the hospital with, are, among others,
fully captured in the dataset, with multiple datapoints for each ranging from mild
to complicated to severe.


#### Tasks

* `medtext`: Open-ended QA in english.
* `medtext_perplexity`: Open-ended QA in english, evaluated with perplexity.

### Citation

```bibtex
@misc{melamud2019automaticgenerationshareablesynthetic,
      title={Towards Automatic Generation of Shareable Synthetic Clinical Notes Using Neural Language Models},
      author={Oren Melamud and Chaitanya Shivade},
      year={2019},
      eprint={1905.07002},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/1905.07002},
}
```