---
language:
- en
license: apache-2.0
size_categories:
- 1M<n<10M
task_categories:
- text-classification
pretty_name: VAERS outcomes
dataset_info:
  features:
  - name: id
    dtype: int32
  - name: text
    dtype: string
  - name: label
    dtype:
      class_label:
        names:
          '0': No event
          '1': ER_VISIT
          '2': ER_VISIT, HOSPITAL
          '3': DIED
          '4': HOSPITAL
          '5': ER_VISIT, HOSPITAL, DIED
          '6': HOSPITAL, DIED
          '7': ER_VISIT, DIED
  splits:
  - name: train
    num_bytes: 779491384
    num_examples: 1270444
  - name: test
    num_bytes: 166688335
    num_examples: 272238
  - name: val
    num_bytes: 165858410
    num_examples: 272238
  download_size: 558443156
  dataset_size: 1112038129
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
  - split: val
    path: data/val-*
tags:
- medical
---

<!-- Provide a quick summary of the dataset. -->

This data set contains VAERS `SYMPTOM_TEXT` records paired with a list of potential outcomes.


### Source Data

Data originates from [VAERS](https://vaers.hhs.gov). The code including the transformations that generated this dataset is [here](https://github.com/chrisvoncsefalvay/daedra/blob/main/notebooks/Dataset%20preparation.ipynb).

### Encoding

The `label` column now comrpises a single powerset-encoded class.