README.md

metadata

license: cc-by-nc-sa-4.0
tags:
  - text-classification
  - bias-detection
  - media
  - news
  - crowdsourcing
  - gamification
task_categories:
  - text-classification
task_ids:
  - multi-class-classification
language:
  - en
pretty_name: NEWS NINJA
size_categories:
  - n<1K

Dataset Card for NEWS NINJA (Game Study dataset)

Dataset Summary

NEWS NINJA contains sentence-level labels of linguistic media bias collected for the paper
“News Ninja: Gamified Annotation of Linguistic Bias in Online News” (https://doi.org/10.1145/3677092).

The dataset has 520 English news sentences: 370 re-labeled from BABE (first 370 entries) and 150 newly labeled sentences (last 150). Each record includes the sentence text, an identifier, a binary majority-vote label, and word-level bias highlights. Single-annotator labels are included as separate columns.

GitHub: https://github.com/Media-Bias-Group/News-Ninja/tree/main
Dataset table (CSV): https://github.com/Media-Bias-Group/News-Ninja/blob/main/News%20Ninja%20Dataset/ExportNewsNinja.csv

Use Cases

• Training/evaluating linguistic bias detection models
• Auditing word-/sentence-level bias cues
• Combining with BABE for larger training corpora

Dataset Structure

Data Fields

• sentencePK (string/int): Unique sentence identifier.
• sentences (string): The news sentence text.
• majority_vote (int; 0 or 1): 1 = biased, 0 = not biased.
• words (string/list): Words marked as biased (delimited list in CSV).
• (plus) per-annotator columns with individual labels (one column per annotator).

Citation

@article{hinterreiter2024ninja,
  title     = {News Ninja: Gamified Annotation of Linguistic Bias in Online News},
  author    = {Hinterreiter, Smi and Spinde, Timo and Oberd{\"o}rfer, Sebastian and Echizen, Isao and Latoschik, Marc Erich},
  journal   = {Proceedings of the ACM on Human-Computer Interaction},
  volume    = {8},
  number    = {CHI PLAY},
  articleno = {327},
  year      = {2024},
  doi       = {10.1145/3677092},
  url       = {https://dl.acm.org/doi/10.1145/3677092}
}