---
pretty_name: German Newspaper Pages — Rights Unresolved (WORK IN PROGRESS)
license: other
license_name: rights-status-unresolved
license_link: https://www.deutsche-digitale-bibliothek.de/content/lizenzen
language:
  - de
task_categories:
  - image-to-text
  - object-detection
tags:
  - ocr
  - layout-analysis
  - historical
  - newspapers
  - webdataset
  - orphan-works
  - work-in-progress
size_categories:
  - 100K<n<1M
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/*.tar
---
> ### 📋 On the licensing of this data
>
> **Every item here is labelled to the best of my ability.** The rights statement is taken
> per issue from what the holding institution recorded in the Deutsche Digitale Bibliothek,
> captured at download time — never inferred, and never applied at newspaper level to issues
> that may differ.
>
> Labelling at this scale is imperfect. **If you believe anything here is mislabelled, or you
> hold rights in any of this material, please write to
> [lorenz.hufe@posteo.de](mailto:lorenz.hufe@posteo.de) — I will resolve it immediately.**

> # 🚧 WORK IN PROGRESS — rights status unresolved
>
> **The copyright status of this material has not been determined.** Every issue here
> carries the Deutsche Digitale Bibliothek rights statement
> [`lizenzen/unbekannt`](https://www.deutsche-digitale-bibliothek.de/content/lizenzen)
> — "rights unknown". That is an **absence of a determination by the holding library, not a
> grant of permission and not an assertion of copyright**.
>
> I am actively working to clarify the rights on these pages with the holding institutions.
> Until that is done, please treat this dataset as provisional:
>
> - **Do not assume it is public domain.** It is not labelled as such by anyone.
> - **It is published for research and rights-clarification purposes.** If you reuse it,
>   satisfy yourself that your use is lawful in your jurisdiction.
> - **Rights holders:** if you hold rights in any of this material and want it removed,
>   please open a discussion on this dataset or contact the dataset owner. It will be taken
>   down promptly, no questions asked.
> - The set may be **withdrawn, re-licensed, or partly removed** as clarification proceeds.
>
> Material whose rights *are* cleared as public domain lives in the companion release
> [`ai-historian/german-newspaper-pages`](https://huggingface.co/datasets/ai-historian/german-newspaper-pages)
> (7,000,060 pages, CC0 / PD Mark). **If you just want usable data, use that one.**

> ## ⚠️ Content notice · Inhaltshinweis
>
> **This collection contains National Socialist propaganda and other historical material
> that is antisemitic, racist and otherwise discriminatory. It is published solely as a
> historical source, for research, teaching and documentation.**
>
> The material runs from 1900 to the 1930s and includes the German press under National
> Socialism: state and party propaganda, antisemitic agitation, notices naming persecuted
> individuals, and reporting that legitimised persecution and war. The OCR text reproduces
> this language verbatim.
>
> Reproduction here is documentary and implies no endorsement. German law permits use of
> such material for civic education, research, teaching, art and reporting on history — the
> social-adequacy clause of §§ 86, 86a StGB. Users are responsible for lawful use in their
> own jurisdiction.
>
> **Diese Sammlung enthält nationalsozialistische Propaganda sowie weiteres historisches
> Material antisemitischen und rassistischen Inhalts. Sie wird ausschließlich als
> historische Quelle zu Zwecken der Forschung, der Lehre und der Dokumentation
> bereitgestellt.** Die Wiedergabe erfolgt dokumentarisch und stellt keine Billigung dar.

# German Newspaper Pages — Rights Unresolved

689,804 pages from 83,030 issues of German newspapers, held by libraries that have **not
determined the copyright status**. Same structure, models and processing as the
public-domain release, so the two can be used together once rights are settled.

| | |
|---|---|
| Pages | 689,804 |
| Issues | 83,030 |
| Rights statement | DDB `lizenzen/unbekannt` (unresolved) — **all pages** |
| Period | 1900s–1930s (nothing before 1900) |
| Format | WebDataset tar shards, images at 2400 px long edge, JPEG q82 |
| Scaled down from | ~3.25 TB at original resolution (median long edge 6,115 px) |

## Why this exists separately

The parent crawl covered ~9.16 M OCR'd pages. Filtering strictly per issue on the rights
statement recorded at download time gave 7.0 M public-domain pages (published separately)
and left 2.14 M pages that are not PD. Those split three ways:

| Rights statement | Pages | Published? |
|---|---:|---|
| In Copyright (InC 1.0) | 1,191,749 | no |
| **Rights unknown (this set)** | **689,804** | **yes, provisionally** |
| Copyright Not Evaluated (CNE 1.0) | 257,304 | no |

Only the *unknown* slice is here. "In Copyright" is an explicit assertion and is not
published. "Not Evaluated" is being reviewed separately — much of it is pre-1900 and
very likely public domain under § 66 UrhG, but that needs confirming, not assuming.

## Contents of a sample

Key `<date>_<issuehash>_<pageid>`, as in the public-domain release:

| Member | Always? | What it is |
|---|---|---|
| `.jpg` | yes | page scan, long edge capped at 2400 px, JPEG q82 |
| `.json` | yes | metadata + layout regions with boxes and OCR text |
| `.alto.xml` | some | the library's own ALTO, verbatim, where the crawl got one |

`.json` boxes are in the coordinates of the `.jpg` beside them; `.alto.xml` keeps the
original full-resolution coordinates, with `image.scale_from_original` bridging the two.
Layout from `PP-DocLayout_plus-L`, transcription from `churro-3B`.

## Limitations

- **Not full resolution.** Originals are ~2.5× larger; fetch them from the `ddb_url` in each
  record if you need archival quality.
- **OCR is machine output.** The model loops or truncates on a minority of regions; filter
  on `finish_reason` if clean text matters.
- **Layout boxes are detections**, not verified structure.
- Coverage reflects what happens to be digitised and rights-unresolved, not what was printed.

## Acknowledgements

These pages were digitised and catalogued by German libraries and archives and made
available through the [Deutsche Digitale Bibliothek](https://www.deutsche-digitale-bibliothek.de/).
Each record carries `provider_ddb_id` and `ddb_url` identifying the holding institution —
please credit it. Only the layout boxes and OCR text originate with this project.