Datasets:

MBZUAI
/

temporal-framing

license: other
gated: true
pretty_name: Temporal Framing
task_categories:
  - text-classification
language:
  - en
  - de
size_categories:
  - n<1K
extra_gated_heading: Request access to the Temporal Framing dataset
extra_gated_description: >-
  This dataset contains third-party news text distributed under controlled
  access for research use.
extra_gated_button_content: Agree and request access
extra_gated_prompt: >-
  By requesting access, you agree to use this dataset only for non-commercial
  research and educational purposes, not to redistribute the raw text, and to
  comply with applicable copyright and licensing restrictions.
extra_gated_fields:
  Affiliation: text
  Country: country
  Intended use:
    type: select
    options:
      - Academic research
      - Education
      - Industry research
      - Journalism / media analysis
      - Other
  Project or lab webpage: text
  I agree not to redistribute the raw text: checkbox
  I agree to use this dataset for non-commercial research or education only: checkbox

Temporal Framing Dataset

This repository contains a sentence-level temporal framing corpus introduced in an accompanying ACL 2026 paper (to appear).

Overview

The dataset consists of English and German news articles annotated for persuasive temporal framing at the sentence level. It is designed to support research on temporal framing detection, classification, and analysis across languages.

The corpus includes 458 news articles, with 2,365 sentences containing temporal framing and 3,030 frame annotations across eight temporal frame categories.

Files

`documents.jsonl`

Each line corresponds to a single news article and includes:

document_id: Unique document identifier
text: Full article text
split: Dataset split (train, dev, or test)
language: Language code (e.g., en, de)
source: Media outlet or source name
publication_date: Article publication date

The full text is provided to support contextual and discourse-level analysis.

`annotations.jsonl`

Each line corresponds to a single annotated sentence and includes:

document_id: Identifier linking the sentence to its document
sentence: Sentence text
labels: One or more temporal framing labels assigned to the sentence

Documentation

Detailed dataset documentation, including motivation, collection methodology, annotation procedures, licensing, and ethical considerations, is provided in DATASHEET.md.

Access Conditions

This dataset contains third-party news article text. We do not claim ownership of the original content.

Access is granted strictly for non-commercial research and educational use. By requesting access, users agree:

not to redistribute the raw article text;
not to use the dataset for commercial purposes;
to comply with applicable copyright and licensing restrictions;
to cite the associated paper and dataset in any resulting work.

If you are a rights holder and wish to request content review or removal, please contact the dataset authors.

License

This dataset consists of two components with different licensing terms:

Article Texts
Full-text news articles are provided under controlled access for research purposes only. Copyright remains with the original publishers.
Annotations and Metadata
All annotations and derived metadata are licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Citation

@inproceedings{mahmoud2026temporal,
  title     = {Uncovering Temporal Framing in the News},
  author    = {Mahmoud, Tarek and Solopova, Veronika and Sahitaj, Premtim and Sahitaj, Ariana and Upravitelev, Max and Abassy, Mervat and Shaikh, Hana Fatima and Foroutan, Neda and Schmitt, Vera and Nakov, Preslav},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (ACL)},
  year      = {2026},
  note      = {To appear}
}