ArMPro

This repo contains the Arabic propaganda dataset (ArMPro).

Table of contents:

• Dataset
  • Data splits
  • Coarse-grained label distribuition
  • Fine-grained label distribuition
• Licensing
• Citation

Dataset

This dataset represents the largest one to date for fine-grained propaganda detection for Arabic. It includes 8,000 paragraphs extracted from over 2,800 Arabic news articles, covering a large variety of news domains.

Example annotated paragraph:

Data splits

We split the dataset in a stratified manner, allocating 75%, 8.5%, and 16.5% for training, development, and testing, respectively. During the stratified sampling, the multilabel setting was considered when splitting the dataset. This ensures that persuasion techniques are similarly distributed across the splits.

Coarse-grained label distribuition

Binary label	Train	Dev	Test
Propagandistic	3,777	425	832
Non-Propagandistic	2,225	247	494
Total	6,002	672	1,326

Coarse-grained label	Train	Dev	Test
Manipulative Wording	3,460	387	757
no technique	2,225	247	494
Reputation	1,404	163	314
Justification	471	48	102
Simplification	384	42	82
Call	176	21	40
Distraction	74	9	16
Total	8,194	917	1,805

Fine-grained label distribuition

Technique	Train	Dev	Test
Loaded Language	7,862	856	1670
no technique	2,225	247	494
Name Calling/Labeling	1,526	158	328
Exaggeration/Minimisation	967	113	210
Questioning the Reputation	587	58	131
Obfuscation/Vagueness/Confusion	562	62	132
Causal Oversimplification	289	33	67
Doubt	227	27	49
Appeal to Authority	192	22	42
Flag Waving	174	22	41
Repetition	123	13	30
Slogans	101	19	24
Appeal to Fear/Prejudice	93	11	21
Appeal to Hypocrisy	82	9	17
Consequential Oversimplification	81	10	19
False Dilemma/No Choice	60	6	13
Conversation Killer	53	6	13
Appeal to Time	52	6	12
Appeal to Popularity	44	4	8
Appeal to Values	38	5	9
Red Herring	38	4	8
Guilt by Association	22	2	5
Whataboutism	20	4	4
Straw Man	19	2	4
Total	15,437	1,699	3,351

Note: "no technique" refers to paragraphs without any propagandistic techniques use.

Licensing

This dataset is licensed under CC BY-NC-SA 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-nc-sa/4.0/

Citation

If you use our dataset in a scientific publication, we would appreciate using the following citations:

• Maram Hasanain, Fatema Ahmad, and Firoj Alam. 2024. Can GPT-4 Identify Propaganda? Annotation and Detection of Propaganda Spans in News Articles. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 2724–2744, Torino, Italia. ELRA and ICCL.

@inproceedings{hasanain-etal-2024-gpt,
    title = "Can {GPT}-4 Identify Propaganda? Annotation and Detection of Propaganda Spans in News Articles",
    author = "Hasanain, Maram  and
      Ahmad, Fatema  and
      Alam, Firoj",
    editor = "Calzolari, Nicoletta  and
      Kan, Min-Yen  and
      Hoste, Veronique  and
      Lenci, Alessandro  and
      Sakti, Sakriani  and
      Xue, Nianwen",
    booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",
    month = may,
    year = "2024",
    address = "Torino, Italia",
    publisher = "ELRA and ICCL",
    url = "https://aclanthology.org/2024.lrec-main.244",
    pages = "2724--2744",
    abstract = "The use of propaganda has spiked on mainstream and social media, aiming to manipulate or mislead users. While efforts to automatically detect propaganda techniques in textual, visual, or multimodal content have increased, most of them primarily focus on English content. The majority of the recent initiatives targeting medium to low-resource languages produced relatively small annotated datasets, with a skewed distribution, posing challenges for the development of sophisticated propaganda detection models. To address this challenge, we carefully develop the largest propaganda dataset to date, ArPro, comprised of 8K paragraphs from newspaper articles, labeled at the text span level following a taxonomy of 23 propagandistic techniques. Furthermore, our work offers the first attempt to understand the performance of large language models (LLMs), using GPT-4, for fine-grained propaganda detection from text. Results showed that GPT-4{'}s performance degrades as the task moves from simply classifying a paragraph as propagandistic or not, to the fine-grained task of detecting propaganda techniques and their manifestation in text. Compared to models fine-tuned on the dataset for propaganda detection at different classification granularities, GPT-4 is still far behind. Finally, we evaluate GPT-4 on a dataset consisting of six other languages for span detection, and results suggest that the model struggles with the task across languages. We made the dataset publicly available for the community.",
}