linksunten.indymedia.org (archive)

73 970 posts from linksunten.indymedia.org, the German-language open-posting platform of the Indymedia network that was banned in 2017, taken from a static JSON export of the site's archive and published as a Hugging Face dataset. This is part of the German Media collection, a set of comparable corpora covering the German media landscape from public-service broadcasters to alternative and fringe outlets. Unlike most datasets in the collection, this one was not crawled but consolidated from an archive export of the site; its columns therefore differ from the crawled news datasets.

⚠️ For scientific research only

This corpus is published exclusively for scientific research under the German research exception for text and data mining — the Forschungsprivileg of § 60d UrhG (implementing Art. 3 of EU Directive 2019/790). The posts remain the copyright of their individual authors and are not licensed for commercial use, redistribution, or republication. By downloading it you confirm that you use it for non-commercial scientific research. See Licensing and intended use.

The posts are reproduced verbatim from the archive of the site; their content expresses the views of their (mostly anonymous) authors, not of the dataset author.

At a glance

Source	linksunten.indymedia.org (offline since August 2017), via a static JSON export of its archive
Rows	73 970
Columns	18 (nid, type, language, tnid, uid, author, created, title, images, filename, oid, tags_themen, tags_orte, tags_regionen, tags_ticker, teaser, body, content)
Language	mostly German: de 93.1 %, en 4.7 %, fr 1.1 %, it 0.7 %, es 0.3 %, nl < 0.1 %
Split	train (single split)
Format	Parquet, 2 shards, ≈ 414 MB download / 697 MB in memory
Perceived partisanship	2.0 / 7 → left of centre (medienkompass.org reader survey, entry indymedia)
Intended use	Scientific research only (Forschungsprivileg, § 60d UrhG)
Mirror	NLP-UniBW/linksunten

from datasets import load_dataset

ds = load_dataset("SinclairSchneider/linksunten", split="train")
print(ds[0]["title"])

About the source

linksunten.indymedia.org was a German-language regional branch of the global Indymedia network (Independent Media Center, founded in 1999). Online from 2008/2009 and run by a collective based in Freiburg im Breisgau, it worked on the open-posting principle: anyone could publish texts, images and event announcements without registering, usually anonymously, while a moderation collective removed posts that broke the site's rules. It became one of the main communication channels of the German left-wing and radical-left movement, carrying reports from demonstrations, campaign calls, an event calendar and anti-fascist research, as well as claims of responsibility for militant actions.

On 25 August 2017 the Federal Ministry of the Interior banned "linksunten.indymedia" as an association under the Associations Act (Vereinsgesetz), calling it the most influential platform of violence-oriented left-wing extremists in Germany, and the site was taken offline. The Federal Administrative Court (Bundesverwaltungsgericht) dismissed the actions brought against the ban in January 2020. The Federal Office for the Protection of the Constitution (Bundesamt für Verfassungsschutz) classified the platform as left-wing extremist.

Legal context. Because the platform was banned as an association, republishing its content can raise legal questions in Germany beyond copyright. Keep the corpus within the research context described under Licensing and intended use, and check with your institution before quoting at length or redistributing material from it.

Perceived political position

The German media compass at medienkompass.org places German mainstream and alternative outlets on two axes on the basis of a reader survey (n ≈ 1 154 participants in total, building on a media map by the think tank polisphere; the quality axis is adapted from Vanessa Otero's media chart).

• x axis — Parteilichkeit (partisanship), 1–7: < 3.5 left, 3.5 – 4.5 relatively neutral, > 4.5 right. 1 far-left (fake news and conspiracy theories) · 2 left mission (questionable journalistic standards) · 3 leaning left · 4 minimal partisan tendency · 5 leaning right · 6 right mission (questionable journalistic standards) · 7 far-right (fake news and conspiracy theories).
• y axis — Qualität (quality), 1–5: 1 sensationalist / clickbait · 2 basic information · 3 meets high standards · 4 analytical · 5 complex.

The survey has a single entry, indymedia, that does not distinguish linksunten from the other Indymedia sites; the scores below are those of that entry.

In that survey, indymedia is rated:

	Score	Reading
Partisanship (1–7)	2.0	left of centre
Quality (1–5)	1.7	on the 1 = clickbait … 5 = complex scale

Parteilichkeit (partisanship):  1 = far left  ...  7 = far right

1         2         3         4         5         6         7
|---------|---------|---------|---------|---------|---------|
          ^                                                  
      indymedia = 2.0

left  <--------------|  neutral  |-------------->  right
                   3.5           4.5

For context, here is the full survey table, sorted from left to right — the other outlets in the collection are directly comparable:

Outlet	Partisanship (1–7)	Quality (1–5)
indymedia ⬅ this dataset	2.0	1.7
Jungle World	2.3	2.3
junge Welt	2.4	2.3
neues Deutschland	2.6	2.3
der Freitag	2.7	2.9
die Tageszeitung	2.8	2.7
Vice	2.8	1.8
NachDenkSeiten	3.1	3.0
Frankfurter Rundschau	3.4	2.8
SPON	3.5	2.1
Spiegel Print	3.5	2.9
Süddeutsche	3.5	3.3
WDR	3.5	2.8
Tagesspiegel	3.6	2.8
Zeit	3.6	3.5
ARD	3.7	2.9
NDR	3.7	2.8
Deutschlandfunk	3.8	3.5
Stern	3.8	2.1
ZDF	3.8	2.8
MDR	4.1	2.5
Handelsblatt	4.3	3.3
NTV	4.3	2.1
BR	4.4	2.6
N24	4.4	2.0
Wirtschaftswoche	4.4	3.1
Frankfurter Allgemeine	4.5	3.3
RTL	4.5	1.5
Die Welt	4.8	2.5
Cicero	4.9	3.1
Focus Online	4.9	1.7
Deutsche Wirtschafts Nachrichten	5.0	2.2
MM News	5.1	1.7
RT Deutsch	5.1	2.0
Achgut	5.2	2.4
BILD	5.2	1.3
Tichys	5.5	2.6
junge Freiheit	5.8	1.9
Compact	6.0	1.8

How to read these numbers. They are perception data from a self-selected online survey, not a content analysis: participants rated outlets they know, the sample is not representative of the German population, and the site itself describes the project as an early iteration. Treat them as a rough, crowd-sourced orientation — useful for grouping datasets into coarse political blocks, not as ground truth for bias labelling.

Dataset structure

Single train split with 73 970 rows and the following columns:

Column	Type	Description	Extracted from
nid	int64	Drupal node id of the post on the original site.	nid of json/node/<nid>.json
type	string	Content type: story (46 %), crossposting (repost, 29 %), event (calendar entry, 23 %), upload, page, image, audio, data, rumour.	type
language	string	Language of the post as set on the site.	language
tnid	int64	Translation-set id linking translated versions of the same post; 0 = none.	tnid
uid	int64	Numeric id of the posting account; 0 = anonymous posting (86 % of rows).	uid
author	string	Free-text author name entered by the poster, usually a pseudonym; empty for 30 % of rows.	author
created	int64	Creation time as a Unix timestamp (seconds, UTC).	created
title	string	Title of the post.	title
images	list[int64]	Node ids of images attached to the post; the images themselves are not included.	images
filename	string	File name of an attached upload (95 rows).	filename
oid	string	Numeric object id carried over from the archive export (78 rows).	oid
tags_themen	list[string]	Topic tags (Themen).	tags.Themen
tags_orte	list[string]	Place tags (Orte).	tags.Orte
tags_regionen	list[string]	Region tags (Regionen).	tags.Regionen
tags_ticker	list[string]	Ticker tags (Ticker).	tags.Ticker
teaser	string	Teaser of the post, as HTML.	teaser
body	string	Full post body as HTML, as stored in the archive.	body
content	string	Plain-text / Markdown rendering of body.	html2text rendering of body, added after the notebook (see below)

Example record

{
  "nid": 221566,
  "type": "crossposting",
  "language": "de",
  "tnid": 0,
  "uid": 0,
  "author": "",
  "created": 1503645952,
  "title": "Innenministerium verbietet linkes Nachrichtenportal \"linksunten.indymedia.org\"",
  "images": [],
  "filename": "",
  "oid": "",
  "tags_themen": [
    "Medienaktivismus & Internet",
    "Repression"
  ],
  "tags_orte": [],
  "tags_regionen": [
    "D-linksunten",
    "Deutschland"
  ],
  "tags_ticker": [],
  "teaser": "<p>Es ist ein schwerer Angriff auf die linke Bewegung und die Pressefreiheit: Heute früh wurde das Portal \"linksunten.indymedia.org\" verboten.\r\nLaut Medienberichten wurde eine entsprechende Verbotsverfügung den mutmaßlichen Betreibern bei drei Hausdurchsuchungen in Freiburg zugestellt. Demnach richte sich die Website gegen die \"verfassungsmäßige Ordnung\" und laufe „nach Zweck und Tätigkeit den Str …",
  "body": "<p>Es ist ein schwerer Angriff auf die linke Bewegung und die Pressefreiheit: Heute früh wurde das Portal \"linksunten.indymedia.org\" verboten.\r\nLaut Medienberichten wurde eine entsprechende Verbotsverfügung den mutmaßlichen Betreibern bei drei Hausdurchsuchungen in Freiburg zugestellt. Demnach richte sich die Website gegen die \"verfassungsmäßige Ordnung\" und laufe „nach Zweck und Tätigkeit den Str …",
  "content": "Es ist ein schwerer Angriff auf die linke Bewegung und die Pressefreiheit:\nHeute früh wurde das Portal \"linksunten.indymedia.org\" verboten. Laut\nMedienberichten wurde eine entsprechende Verbotsverfügung den mutmaßlichen\nBetreibern bei drei Hausdurchsuchungen in Freiburg zugestellt. Demnach richte\nsich die Website gegen die \"verfassungsmäßige Ordnung\" und laufe „nach Zweck\nund Tätigkeit den Strafge …"
}

Temporal coverage

The created column could be parsed for 73 970 of 73 970 rows (100.0 %). Those rows span 2008-02-13 → 2017-08-25.

Year	Posts
2008	4	█
2009	3 157	██████
2010	4 242	████████
2011	6 648	████████████
2012	7 266	█████████████
2013	9 055	█████████████████
2014	9 856	██████████████████
2015	11 928	██████████████████████
2016	13 136	████████████████████████
2017	8 678	████████████████

How the data was collected

This dataset was not crawled. linksunten.indymedia.org has been offline since August 2017 (see About the source). The corpus was built from a static JSON export of the site's archive, which has one JSON file per Drupal object in folders such as node/, comment/, user/, image/, themen/ and orte/, all dated January 2020. Only json/node/ (193 324 posts) was used. Comments, user records and media files are not part of the dataset.

Consolidation — scraping/finalize.ipynb

1. glob("json/node/*.json") collects every post (193 324 files).
2. Each record is flattened into one pandas.DataFrame column per field; the tags dictionary is split into the four list columns tags_themen, tags_orte, tags_regionen and tags_ticker.
3. Posts without text are removed: only rows where html2text.html2text(body).strip() != '' are kept (74 111 rows). This drops pure event stubs, image and file nodes without a description, and similar.
4. drop_duplicates(subset=['body']) removes identical bodies (73 970 rows).
5. The frame is written to linksunten.json / .csv and pushed with datasets.Dataset.from_pandas(df).push_to_hub("SinclairSchneider/linksunten").

The content column

The notebook produces 17 columns. The content column on the Hub was added in a later step that is not part of the notebook. It holds a Markdown-style plain-text rendering of body made with html2text (emphasis as _…_, lines wrapped at about 78 characters, runs of blank lines collapsed). Running html2text.html2text(body).strip() on body reproduces it closely; the exact whitespace and some formatting details depend on the html2text version.

Reproducing the consolidation

pip install pandas datasets huggingface_hub html2text tqdm

# place the archive export's json/ folder next to the notebook, then
jupyter notebook finalize.ipynb

The archive export itself is not distributed with this repository.

Usage

from datasets import load_dataset
import pandas as pd

ds = load_dataset("SinclairSchneider/linksunten", split="train")

# stream instead of downloading everything
stream = load_dataset("SinclairSchneider/linksunten", split="train", streaming=True)
print(next(iter(stream))["title"])

# as a DataFrame
df = ds.to_pandas()
df["created"] = pd.to_datetime(df["created"], unit="s", utc=True)
recent = df[df["created"].dt.year >= 2016]
print(len(recent), "posts from 2016 onwards")

Combining several outlets into one comparative corpus:

from datasets import load_dataset, concatenate_datasets

outlets = ["SinclairSchneider/linksunten", "SinclairSchneider/tagesschau_de"]
parts = []
for name in outlets:
    d = load_dataset(name, split="train").select_columns(["title", "content"])
    parts.append(d.add_column("outlet", [name.split("/")[-1]] * len(d)))
corpus = concatenate_datasets(parts)

Limitations and biases

• Snapshot of a banned platform. The corpus contains what the 2020 archive export held: posts removed by the moderators before August 2017 are missing, and the timeline ends abruptly on 25 August 2017, the day of the ban. Only a handful of posts predate 2009.
• Open posting, not journalism. Anyone could publish, usually anonymously. Posts range from event announcements and demonstration reports to manifestos and claims of responsibility for militant actions, and 29 % are reposts (crossposting). Nothing was fact-checked or edited.
• Personal data. Posts can name or describe private individuals (for example research articles about alleged far-right activists), and author / uid identify posting accounts. Texts about political views are special-category data under Art. 9 GDPR. Process the corpus under your institution's research data safeguards, do not try to re-identify posters, and do not republish personal details.
• Not only German. About 7 % of posts are in English, French, Italian, Spanish or Dutch; filter on language if you need a monolingual corpus.
• Deduplication by body. Identical bodies were collapsed; near-duplicates (the same text reposted with small edits, or a post and its translation linked via tnid) remain.
• HTML in body and teaser. These are raw HTML; use content for plain text.
• Movement perspective. The platform was used almost exclusively by the left-wing and radical-left scene. Any model trained on it alone inherits that milieu's topics, framing and vocabulary; combine it with other datasets of the collection for comparative work.

Licensing and intended use

Scientific research only — Forschungsprivileg (§ 60d UrhG)

This dataset is made available solely for the purpose of scientific research. The legal basis is the German text-and-data-mining exception for research, the so-called Forschungsprivileg:

• § 60d UrhG — Text und Data Mining für Zwecke der wissenschaftlichen Forschung — allows research organisations to make reproductions of lawfully accessible works in order to carry out text and data mining for non-commercial scientific research, and to share the resulting corpus with a specifically limited circle of people for joint research and with individual third parties for the purpose of quality assurance.
• It implements Art. 3 of Directive (EU) 2019/790 (the DSM Directive); comparable research exceptions exist in the other EU member states.
• § 60d UrhG does not cover commercial exploitation. Research organisations that give a commercial enterprise preferential access are explicitly excluded from the privilege.

What this means for you

✅ Permitted — using the corpus for non-commercial scientific research: text and data mining, corpus linguistics, media-bias and discourse analysis, training and evaluating research models, reproducing published findings, and quality checks by individual third parties.

❌ Not permitted — commercial use or commercial model training; redistributing or republishing the article texts; presenting the texts as your own; building products or services on the article texts; any use that substitutes for reading the publisher's own website.

The posts remain the copyright of their individual, mostly anonymous authors. No licence to them is granted or implied by this repository. Only the consolidation code in scraping/ is offered for free reuse.

If your intended use falls outside the research exception, obtain permission from the rights holder, or use only the derived, non-expressive data (statistics, embeddings, counts). Note that § 60d UrhG also expects research corpora to be deleted once they are no longer needed for the research purpose, with archiving permitted at trusted institutions.

If you are a rights holder and want content removed, please open a discussion on this repository.

This section is an orientation for researchers, not legal advice. Responsibility for your own use of the data remains with you and your institution.

Citation

@misc{linksunten_dataset,
  title        = {linksunten.indymedia.org (archive): a corpus of posts from the archive of the open-posting platform},
  author       = {Schneider, Sinclair},
  year         = {2023},
  howpublished = {Hugging Face},
  url          = {https://huggingface.co/datasets/SinclairSchneider/linksunten}
}

See also

• The full German Media collection
• Mirror of this dataset: NLP-UniBW/linksunten
• medienkompass.org — Deutsche Medienlandschaft, the source of the perceived-partisanship figures above