Datasets:
language:
- da
license: cc-by-4.0
size_categories:
- 1M<n<10M
pretty_name: Folketinget Denmark with Blame Labels
task_categories:
- text-classification
tags:
- politics
- debates
dataset_info:
features:
- name: text
dtype: large_string
- name: labels
dtype: int64
- name: speaker
dtype: large_string
- name: date
dtype: large_string
- name: party
dtype: large_string
- name: paragraph_nr
dtype: int64
- name: sentence_nr
dtype: int64
splits:
- name: train
num_bytes: 103096732
num_examples: 450000
- name: test
num_bytes: 100077
num_examples: 424
- name: validation
num_bytes: 11440803
num_examples: 50000
- name: inference
num_bytes: 1167411515
num_examples: 5098494
download_size: 467309657
dataset_size: 1282049127
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
- split: validation
path: data/validation-*
- split: inference
path: data/inference-*
Label Scheme:
Blame: 1
No Blame: 0
This dataset was created partially from The ParlSpeech V2 data set (Rauh, C., & Schwalbach, J. (2020)) from 07/10-1997 up to and including 20/12-2018, with data from 08/01-2019 up to and including 26/02-2026 coming from our own fetch from Folketingets own sFTP server.
Blame labels were created on a machine translated version of the data (using opus-MT) by the Political DEBATE model.
The train / validation split is a simple 90/10 split of the machine labelled data, with the test set being manually annotated by two authors with an inter-annotator agreement of 84.8% - a Cohen's Kappa of 0.675.
Inference data is what remained after cleaning, as described in our arXiv paper, and extraction of training data.
It is possible to recreate the full dataset without splits by joining across the paragraph and sentence number columns, as all data is present.
More details can be found in the GitHub for the BlameBERT Project
If you use this dataset in any of your own research, please cite us:
@misc{jensen2026blamingaislepoliticalcontrasting,
title={Blaming Across the Aisle: Political Contrasting and Blame Attribution in the Danish Parliament},
author={Markus Lundsfryd Jensen and Rune Egeskov Trust and Kenneth Christian Enevoldsen and Sara Kolding},
year={2026},
eprint={2609.26346},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2609.26346},
}