---
dataset_info:
  features:
  - name: id
    dtype: int64
  - name: image
    dtype: image
  - name: conversation
    struct:
    - name: data
      list:
      - name: answer
        dtype: string
      - name: question
        dtype: string
  splits:
  - name: train
    num_bytes: 59333908
    num_examples: 700
  download_size: 59207371
  dataset_size: 59333908
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
task_categories:
- question-answering
language:
- en
tags:
- medical
size_categories:
- n<1K
---

This dataset is made using raw data from [MURA (musculoskeletal radiographs)](https://arxiv.org/abs/1712.06957), a large dataset of bone X-rays.

For each image in the original raw dataset we used the associated metadata to generate a simulated conversation about the image between a user and a chatbot.