GPT-NL/dutch-proverbs

---
dataset_info:
features:
- name: scenario
dtype: string
- name: correct_proverb
dtype: string
- name: incorrect_proverb
dtype: string
splits:
- name: train
num_bytes: 7299
num_examples: 32
- name: test
num_bytes: 23286
num_examples: 98
download_size: 27103
dataset_size: 30585
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
---
# Dutch proverbs

This a collection of 130 brief `scenario's` in Dutch, combined with a proverb (_spreekwoord_) that matches the scenario (`correct spreekwoord`) and a proverb that does not match the scenario (`incorrect spreekwoord)`.

## Creation
The dataset was created and reviewed manually by a team of NFI employees (Netherlands Forensic Institute).

The creation process was as follows:

1. A proverb was selected
2. A scenario was created to match the proverb
3. An alternative, incorrect proverbs was selected

The alternative proverbs were **not** selected randomly. They were crafted to share similar words or a similar theme with the scenario, yet are still clearly the incorrect proverb for the situation.

## Use
These proverbs can be used to test the ability of Large Language Models to match the right proverb to a scenario.