Dataset Viewer
The dataset viewer is taking too long to fetch the data. Try to refresh this page.
Server-side error
Error code:   ClientConnectionError

PoliTok-DE

A Multimodal Dataset of Political TikToks and Deletions From Germany.

Paper: PoliTok-DE: A Multimodal Dataset of Political TikToks and Deletions From Germany

Authors: Tomas Ruiz, Andreas Nanz, Ursula Kristin Schmid, Carsten Schwemmer, Yannis Theocharis, Diana Rieger

Dataset Description

PoliTok-DE contains TikTok post IDs from two German elections:

Collection Window Released post IDs Deleted at re-scrape
Saxony state election 2024 2024-07-01 to 2024-11-30 195,373 36,617 (18.7%)
German federal election 2025 2025-01-13 to 2025-04-30 743,588 295,101 (39.7%)

The released post-ID corpus contains 938,961 posts in total. Posts were identified through the TikTok Research API and can be hydrated with the included scraping code to retrieve currently available multimodal content and metadata from TikTok.

The federal-election release excludes posts outside the well-covered publication window and posts that only matched false-positive blacklisted keywords. Random-sample subsets are not released.

Files

  • data/saxony_2024/post_ids.parquet: Saxony 2024 post IDs.
  • data/federal_2025/post_ids.parquet: Federal election 2025 post IDs.
  • data/saxony_2024/annotations.parquet: Human annotations for a Saxony-only subset.
  • hydrate.py: Hydration script for recollecting currently available post media and metadata.

The post-ID files contain these columns:

  • post_id: TikTok post ID, stored as a string.
  • collection: Collection identifier (saxony_2024 or federal_2025).
  • availability_YYYY_MM_DD: Observed public availability at a re-scrape date.
  • status_codes_YYYY_MM_DD: Pipe-separated TikTok status codes for posts marked DELETED in the matching availability column. Empty otherwise.

Availability values are:

Value Meaning
AVAILABLE The post was successfully re-scraped and was publicly retrievable at that date.
DELETED The post was not publicly retrievable and TikTok returned a durable unavailable, private, or moderation status.
UNKNOWN_ERROR The re-scrape was attempted, but failed with a transient or inconclusive error, such as a WAF block, HTTP block, timeout, or parsing failure.
NOT_RESCRAPED This post was not part of that re-scrape wave.

The Saxony collection has availability snapshots for 2024-10-02, 2024-12-10, and 2025-01-13. The federal-election collection has an availability snapshot for 2026-06-11. Each post-ID file is exposed as a separate Hugging Face config, so each collection keeps only the availability columns that apply to it.

Loading With Hugging Face Datasets

Install the Hugging Face datasets library:

pip install datasets

Load the Saxony 2024 post-ID release:

from datasets import load_dataset

saxony_ids = load_dataset("tomasruiz/PoliTok-DE", "saxony_2024_post_ids")
saxony_ids = saxony_ids["train"]

Load the federal-election 2025 post-ID release:

from datasets import load_dataset

federal_ids = load_dataset("tomasruiz/PoliTok-DE", "federal_2025_post_ids")
federal_ids = federal_ids["train"]

Load the Saxony human annotations:

from datasets import load_dataset

annotations = load_dataset("tomasruiz/PoliTok-DE", "saxony_2024_annotations")
saxony_annotations = annotations["train"]

Hydration Code

The script hydrate.py downloads currently available TikTok content from one post-ID parquet file. It creates:

  1. A video or image and audio file in the media/ directory.
  2. An entry in media_progress.csv, stating whether the post was successfully scraped.
  3. An entry with post metadata in media_metadata.csv, only if the post was successfully scraped.

Install dependencies:

pip install -r requirements.txt

Hydrate the Saxony 2024 collection:

python hydrate.py --post-ids-file data/saxony_2024/post_ids.parquet --output-dir saxony_hydration

Hydrate the federal-election 2025 collection:

python hydrate.py --post-ids-file data/federal_2025/post_ids.parquet --output-dir federal_hydration

Note: To interrupt the script, you might need to press Ctrl+C multiple times due to the concurrent scraping code.

Dataset Access

This repository contains post IDs and Saxony annotations. It does not publicly release raw videos, images, audio, or full metadata. Full content can be hydrated from TikTok when the post is still publicly available. Access to stored deleted content is restricted and can be requested for academic research purposes.

Citation

@article{ruiz2025politok,
  title={PoliTok-DE: A Multimodal Dataset of Political TikToks and Deletions From Germany},
  author={Ruiz, Tomas and Nanz, Andreas and Schmid, Ursula Kristin and Schwemmer, Carsten and Theocharis, Yannis and Rieger, Diana},
  journal={arXiv preprint arXiv:2509.15860},
  year={2025}
}
Downloads last month
50

Paper for tomasruiz/PoliTok-DE