---
task_categories:
- visual-question-answering
language:
- de
tags:
- vera
- math
- vlm
- school
pretty_name: SEGData (Standardised Educational German Data)
size_categories:
- n<1K
dataset_info:
  features:
  - name: image
    dtype: image
  - name: description
    dtype: string
  - name: text
    dtype: string
  - name: answers
    list: string
  - name: gold_answer
    dtype: string
  - name: id
    dtype: string
  - name: question_type
    dtype: string
  - name: class
    dtype: string
  - name: added_multiplechoice
    dtype: bool
  - name: source
    dtype: string
  splits:
  - name: train
    num_bytes: 27789550
    num_examples: 395
  download_size: 27681251
  dataset_size: 27789550
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---


# SEGData (Standardised Educational German Data)

The original data is designed by the _Institut zur Qualitätsentwicklung im Bildungswesen_ to test German school children in classes 3, 8 and 12.
It is freely accessible at their [website](https://www.iqb.hu-berlin.de/de/schule/aufgaben/).

The licence is specific to the task and image.
Most are accessible under the CC BY licence.
See the _source_ column for more information.


# Dataset preprocessing

Only a subset of the available data from the original test is used for this dataset.
First, only tasks of classes 3 and 8 were chosen and only the category "Testaufgaben".

All tasks were removed that

  * contain no image
  * could be solved without an image
  * requires image generation for solving the task

All tasks were transformed into a multiple-choice format. For cloze tests and reasoning tasks, answering options were generated and their order was randomized.