This github provides supplementary materials for our EMNLP-2025 paper:
"Moral Framing in Politics (MFiP): A new resource and models for moral framing"
@inproceedings{rehbein-etal-2025-moral-framing,
title = "Moral Framing in Politics ({MF}i{P}): A new resource and models for moral framing",
author = "Rehbein, Ines and
Reinig, Ines and
Ponzetto, Simone Paolo",
editor = "Christodoulopoulos, Christos and
Chakraborty, Tanmoy and
Rose, Carolyn and
Peng, Violet",
booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.emnlp-main.1757/",
doi = "10.18653/v1/2025.emnlp-main.1757",
pages = "34643--34663",
ISBN = "979-8-89176-332-6",
}
The repository includes the data and annotation guidelines for the MFiP corpus.
In the data folder are two subfolders. The first one, mfip, includes the annotation layers in a json format, with labels annotated by the individual coders and the majority label (see paper for details).
The second folder, xval, includes the preprocessed data used in the crossvalidation experiments described in the paper.
- only used to extract data for data augmentation
- no experimental results reported in the paper (only baseline in appendix, see Appendix D, Moral frame identification (FI))
| Frame type distribution |
|---|
| 45 ImmoralValue |
| 475 MoralValue |
| 1,090 NoFrame |
| 1,314 PoliticalActOrGoal |
| 2,262 ImmoralActOrGoal |
| 2,386 MoralActOrGoal |
| 7,572 Total |
cd MFiP/experiments/data/xval
- create train/test sets for 5-fold cross-validation
bash make_train_test.sh
python stats.py
| Party | # Tokens | # Frames | MoralValue | MoralActOrGoal | ImmoralValue | ImmoralActOrGoal | PoliticalActOrGoal | NoFrame |
|---|---|---|---|---|---|---|---|---|
| AfD | 71275 | 1193 | 68 | 272 | 8 | 541 | 157 | 147 |
| CDU | 161447 | 2218 | 145 | 747 | 6 | 467 | 446 | 407 |
| FDP | 47565 | 799 | 46 | 282 | 1 | 204 | 153 | 113 |
| GRUENE | 40956 | 653 | 53 | 180 | 10 | 222 | 115 | 73 |
| LINKE | 46308 | 868 | 36 | 220 | 2 | 396 | 107 | 107 |
| SPD | 119938 | 1771 | 123 | 664 | 16 | 403 | 324 | 241 |
| fraktionslos | 6562 | 70 | 4 | 21 | 2 | 29 | 12 | 2 |
Please also see our repo with the final release of the corpus, including unified annotation layers for the parliamentary debates (NER, Situation Entities, Speech Events, MoPE annotations, and moral frames/roles). The corpus includes all annotation layers in a unified json format:
Github: https://github.com/umanlp/gepadeu.git
You can also search a pre-final version of the corpus online (w/o roles, to be added soon). The corpus is available via the ANNIS platform, an open source web browser-based search and visualisation architecture for complex multilevel linguistic corpora with diverse types of annotation. See here for more info on ANNIS.
GePaDeU in ANNIS: Search the corpus online.
Paper:
@inproceedings{rehbein-etal-2026-gepadeu,
title = "{G}e{P}a{D}e{U} - a Multi-layer Corpus of {G}erman Parliamentary Debates with Rich Semantic and Pragmatic Annotations",
author = "Rehbein, Ines and
Schlenker, Julian and
Ostertag, Lars and
Ponzetto, Simone Paolo",
editor = "Piperidis, Stelios and
Bel, N{\'u}ria and
van den Heuvel, Henk and
Ide, Nancy and
Krek, Simon and
Toral, Antonio",
booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resource Association",
url = "https://aclanthology.org/2026.lrec-1.141/",
doi = "10.63317/5d28ibqd2nzk",
pages = "1794--1810",
}