A World in Print: Introducing a Danish-Norwegian corpus of historical newspapers

Camilla B{\o}geskov; Johan Heinsen

REVIEW 3 major objections 5 minor 5 references

A World in Print: Introducing a Danish-Norwegian corpus of historical newspapers

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims a neural pipeline turns illegible fraktur microfilm into 474 million words of searchable Danish history.

desk verdict A genuinely useful corpus with an honest write-up, but the headline accuracy number is not yet backed by independent validation. read the letter →

arxiv 2509.02356 v1 pith:N5UNGOKE submitted 2025-09-02 cs.DL

Johan Heinsen , Camilla B{o}geskov This is my paper · ORCID

classification cs.DL

keywords historicalnewspapersDanish-NorwegiancorpusopticalcharacterrecognitionfrakturtypefacetextsegmentationdigitalhumanitiesAbsolutismeradatadescriptor

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Most Danish newspapers printed before 1900 were photographed on microfilm decades ago, then run through standard text-recognition software that produced roughly 50% word accuracy—too broken to search. This paper reports a neural text-recognition pipeline that reads the same microfilm with far higher accuracy, and uses it to reconstruct about 474 million words from 28 Danish and Norwegian newspapers published between 1660 and 1849. The authors state this is by far the largest fully searchable text dataset in historical Danish before 1900, dwarfing the previous best of about 60 million words. Alongside the transcribed text, the dataset provides standalone texts split from the page layouts and a per-text word-accuracy score, so users can filter out passages where recognition failed. The dataset is released with its models, and the authors argue it creates a new foundation for computational history of Denmark-Norway.

What carries the argument

The central machinery is the recognition and segmentation pipeline: (1) a neural text-recognition model trained on about 420,000 words of fraktur print (the black-letter typeface used in Danish newspapers) and intermittent antiqua, which transcribes microfilm scans; (2) a line-classification model that combines structural features—capitalisation, line length, punctuation—with contextual predictions from a masked-language model to decide where a text begins and ends; (3) a dictionary-based word-accuracy indicator that estimates recognition quality per text without ground truth, letting users compare and filter series.

What would settle it

An independent, stratified sample of pages from the under-validated corners—Norwegian papers, warped or low-contrast scans, capitalised-antiqua advertisements—transcribed by a human reader and compared with the model output character-by-character would settle the quality claim. If the independent character error rate is substantially above 0.6%, the 'fully searchable' description does not hold for the whole corpus.

Watch

Extended reading notes

Core claim

The dataset's central claim is that the ENO corpus—474 million words from about 565,000 pages across 28 titles—is the largest fully searchable text collection in historical Danish before 1900, and that it is accurate enough to be used as text. The authors report a character error rate of about 0.6% on validation data, a line-level segmentation F1 of 98.9%, and a per-text word-accuracy indicator built from period dictionaries. They also describe a masked-language model trained on an early version of the corpus, used to improve segmentation into standalone texts and, in future, to correct recognition errors.

Load-bearing premise

The claim collapses if the 0.6% character error rate measured on validation pages, which come from the newspapers already in the training data, does not apply to the rest of the corpus—especially Norwegian papers, warped scans, and capitalised antiqua passages, which the authors themselves flag as different.

Editorial extensions

If this is right

The full corpus becomes searchable by any word, phrase, or regular expression, letting historians find mentions of people, goods, prices, events, and ideas across 28 newspapers.
Each text carries a quality score, so users can filter to high-accuracy passages for language modelling and quantitative text analysis; the authors recommend aggressive filtering for such uses.
Because texts are segmented into standalone units rather than whole editions, researchers can track genres such as labour advertisements, runaway notices, and consumption ads over time.
The trained recognition and segmentation models are released with the data, so other institutions with similar fraktur or microfilm scans can apply the same pipeline.
The inclusion of Norwegian papers up to 1814 allows comparisons across the two sides of the Denmark-Norway state within one corpus.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The per-text quality indicator could be reused as a cheap, ground-truth-free quality audit for other historical OCR collections, since it only needs a period dictionary.
If the recognition model generalises beyond the validation set, the same approach may work on other European fraktur newspapers whose scans are of comparable or worse quality; the authors' own caution about Norwegian and warped scans suggests the boundary should be tested explicitly.
A filtered subset of this corpus could serve as training data for historical Danish language models, which in turn could improve search and error correction in the wider digitised archive.
The segmentation-plus-similarity interface tests a workflow in which historians move horizontally from one text to similar texts rather than only keyword search; if it proves useful, similar 'similar-text' layers could be added to other newspaper archives.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know about this paper because it ships a very large, openly available historical newspaper corpus, and because the central accuracy claim needs a skeptical read.

What's new: 474M words from 28 Danish/Norwegian newspapers covering 1660–1849, produced with Transkribus-based neural OCR on microfilm that previous OCR made useless. That is a real contribution. The dataset, models, BERT embeddings, code, and a public search interface are all released, and the integration with the Danish Dynaword corpus is a plus. The authors are also unusually candid about bias and about the corners they expect to be bad—the Norwegian papers from newer photographs and the warped Internet Archive scans of Nyeste Skilderie.

The soft spots are in the validation story. The reported 0.6% character error rate comes from validation transcriptions drawn from the same newspapers used in training. That is not an independent test set. The in-house PWA indicator is built partly from an early version of the corpus itself, so it can catch gibberish but not real-word errors or rare names. The paper acknowledges both of these limitations, but the headline claim of a 'fully searchable' corpus rests on an unverified transfer from validation to the atypical scans. That transfer might hold, but the paper gives no direct evidence. A proper held-out test sampled across all 28 titles and scan types—including the Norwegian and warped pages—would close the gap. The segmentation F1 of 98.9% on 7,000 manually tagged lines looks more credible, though I'd want confirmation that those lines were not part of the 30,000 training tags.

The flaws are in the validation narrative, not in the artifact. The dataset is new, openly licensed, and the per-series quality flags give users a way to hedge their own risk. Historians and computational linguists working on early modern Danish or Norwegian will find this transformative despite the open accuracy question.

Send it to peer review. A good reviewer will push for the external held-out test and for precise wording about what 'fully searchable' means. If you work in this area, cite it and use the corpus—just don't repeat the 0.6% figure as if it were measured on independent ground truth.

Referee Report

3 major / 5 minor

Summary. The paper introduces Enevældens Nyheder Online (ENO), a dataset of approximately 474 million words reconstructed from about 565,000 pages of 28 Danish and Norwegian newspapers from the period 1660–1849. The authors describe a pipeline built on Transkribus neural text-recognition models, a line-level segmentation approach that combines structural features with a BERT-based model, and a word-level predicted-word-accuracy (PWA) indicator. The central claims are that the corpus is 'fully searchable,' that it is 'by far the largest fully searchable text dataset in historical Danish dating from before 1900,' and that the text recognition achieves a 0.6% character error rate. The paper also documents coverage, usage notes, a similar-text recommendation feature, and provides public access to the data and models.

Significance. If the accuracy and coverage claims hold, this is a major infrastructural contribution to Danish and Norwegian historical scholarship, enabling full-text search, text mining, and language modeling on a corpus that was previously largely unusable for computational purposes. The paper is commendable for making the data and models openly available, for providing per-text quality indicators, and for explicitly warning about historical bias in the corpus. The comparison to the next-largest pre-1900 Danish corpus (about 60 million words) makes the potential impact clear. The central quantitative claim, however, rests on a validation set that is not independent of the training data, and the corpus-wide quality indicator is partly constructed from the corpus itself. The significance of the resource is therefore currently clouded by an unverified generalization of the reported error rate.

major comments (3)

[Validation] The reported 0.6% character error rate is computed on 'validation data, made up of additional transcriptions from the newspapers included in the training material.' This is not an independent test set: the same titles and likely the same scan characteristics are over-represented. The authors themselves note that 'some corners of the dataset contain both types of print and scans different from those in the validation data,' specifically the three Norwegian papers and the warped Internet Archive scans of Nyeste Skilderie. Since the headline claims of 'fully searchable' quality and the comparison to a 60M-word corpus depend on corpus-wide accuracy, the paper must either provide a held-out evaluation stratified by title and scan type, or clearly restrict the accuracy claim to the validation subcorpus.
[Validation (PWA indicator)] The PWA dictionary is built in part from 'the 10.000 most common words in an early version of the corpus' that were not already in canonical literature or census sources. This introduces a circularity: the dictionary can recognize the model's own normalized output, but it cannot detect real-word substitutions that happen to be legitimate period words. The paper acknowledges that the method is 'somewhat simple,' but it is then used to support the statement that 'the data is of a high and consistent quality' and to produce the distribution in Figure 5. The PWA indicator is useful as a descriptive relative measure across texts, but it is not a substitute for independent ground-truth evaluation on atypical scans and titles.
[From pages to texts] The segmentation pipeline reports an f1-score of 98.9% on 'a validation dataset of about 7,000 manually tagged lines.' Because the BERT model was trained on 260 million words from the initial segmentation, and the setfit models were trained on a manually tagged dataset of about 30,000 lines, it is not clear whether the 7,000 validation lines are disjoint from both the setfit training data and the BERT pretraining corpus. This matters because the paper itself describes the 'positive feedback loop' in which the initial segmentation informed a better tool. The authors should state explicitly that the validation lines were not used in any training stage and, ideally, report performance broken down by publication and layout type.

minor comments (5)

[Abstract] 'Enevaeldens' is spelled inconsistently with the later 'Enevældens'; the abstract also contains the grammar error 'The contributions details this process.'
[Methods] 'Tronheim' should be 'Trondheim'.
[Validation] The newspaper title appears as both 'Nyeste Skilderie af Kiøbenhavn' and 'Nyeste Skilderie after Kiøbenhavn'; the latter appears to be a typo for 'af'.
[References] Reference 3 contains a typo: 'Digitsed' should be 'Digitised'.
[Validation / Figure 5] Figure 5 would benefit from explicit axis labels and a definition of the plotted quantity (e.g., median PWA per text? per publication?), and the text should state the size and coverage of the dictionary used for the PWA indicator.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the dataset construction and validation are largely self-contained, with minor transparent self-referential elements that do not carry the central claims.

full rationale

The paper's central claims are that ENO contains ~474M words and is the largest fully searchable historical Danish text dataset before 1900, and that its OCR pipeline yields highly accurate text. These are empirical construction and measurement claims, not predictions derived by construction from their inputs. The reported 0.6% character error rate is an observed value on manually transcribed validation data, not a fitted parameter renamed as a prediction. The validation set is admittedly drawn from newspapers included in the training material, and the authors explicitly warn that 'some corners of the dataset contain both types of print and scans different from those in the validation data.' This is a generalizability limitation, not circularity: the error rate is not forced to equal 0.6% by definition. The BERT-based segmentation feedback loop is also not circular, because the final segmentation performance is evaluated against approximately 7,000 manually tagged lines (f1=98.9%), an external human ground truth rather than the BERT model's own output. The PWA indicator is partially constructed from an early version of the corpus ('we combed through the 10.000 most common words in an early version of the corpus'), but the paper repeatedly describes it as 'crude and lightweight' and uses it only for relative comparisons and for flagging inconsistencies, not as the primary evidence for the 'fully searchable' quality claim. Self-citations to the authors' own models and HuggingFace releases are for reproducibility and do not carry the argument. Overall, no step reduces to its inputs by construction; the main risks are external validity of the error estimate and corpus bias, both of which the paper acknowledges. Score 1 reflects the minor self-referential validation elements, not load-bearing circularity.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper does not introduce new theoretical entities or physical constructs. It relies on assumptions about the fidelity of the source images and the representativeness of validation data. The self-referential dictionary for PWA is a notable assumption that affects quality estimates.

assumptions (3)

domain assumption The digitized microfilm and photographed images are faithful representations of the original newspapers.
The entire dataset is derived from these images; any distortion or loss in microfilming affects the output text. This is implied throughout the Methods section.
domain assumption The validation transcriptions are representative of the broader corpus.
The paper reports a 0.6% character error rate based on validation data from the same newspapers used in training, but acknowledges variations in scan type and quality across the corpus. See the Validation section.
domain assumption The dictionary-based PWA is a valid proxy for text recognition accuracy.
The PWA uses a dictionary partly derived from the corpus itself and may not accurately reflect word-level errors. The authors describe it as 'crude and lightweight' in the Validation section.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A World in Print: Introducing a Danish-Norwegian corpus of historical newspapers." pith.science (2026). https://pith.science/paper/N5UNGOKE

@misc{pith2026250902356,
  author       = {Pith},
  title        = {Pith review of: A World in Print: Introducing a Danish-Norwegian corpus of historical newspapers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N5UNGOKE}},
  note         = {Machine review of arXiv:2509.02356}
}

read the original abstract

This Data Descriptor introduces the dataset Enevaeldens Nyheder Online (News during Absolutism Online). The Enevaeldens Nyheder Online (ENO) dataset provides a reconstruction of the contents of major newspapers in Denmark and Norway during the period of Absolutism (1660-1849). The dataset contains approx. 474 million words, created using neural networks designed to process digitised microfilm versions of Danish newspapers as well as a smaller selection of Norwegian publications that were all hitherto illegible for computers. The contributions details this process and its results, including a way to derive standalone texts from the editions, and the accompanying BERT-model trained on a beta-version of the dataset.

Figures

Figures reproduced from arXiv: 2509.02356 by the authors.

**Figure 2.** Coverage over time. From pages to texts One of our aims with the dataset was to explore certain genres of advertisements. These include advertisements relating to specific forms of consumption, advertisements for labour, and runaway notices. By having a broad corpus we wanted to extract texts of relevance on a scale that allowed for investigation of developments over time. However, one obstacle needed to be overcome… view at source ↗

**Figure 3.** Texts per year. ambitions. An edition might contain many texts, and a text might span multiple pages. Thus, we had to devise a way to segment the output into meaningful units. We decided to approach this as a problem of classifying every line of text, since lines were separated in the output of the text recognition process. If one can infer whether a line represents the beginning of a text, it is easy to split editi… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Example pwa-scores. Quartiles. Aarhus Stifts-Tidende. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: Distribution of pwa-scores. 8 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

5 extracted references · 4 canonical work pages

[1]

A World in Print: Introducing a Danish-Norwegian corpus of historical newspapers

A World in Print Introducing a Danish-Norwegian corpus of historical newspapers Johan Heinsen and Camilla Bøgeskov 2025-09-02 Abstract This Data Descriptor introduces the dataset Enevældens Nyheder Online (News during Absolutism Online). The Enevældens Nyheder Online (ENO) dataset provides a reconstruction of the contents of major newspapers in Denmark an...

work page Pith review arXiv 2025
[2]

The disproportionate amount of data from the second half of the period is shown in figure 3, showing the distribution of individual texts pr. year. Figure 2: Coverage over time. From pages to texts One of our aims with the dataset was to explore certain genres of advertisements. These include advertisements relating to specific forms of consumption, adver...

work page 1989
[5]

7 Figure 4: Example pwa-scores

14https://archive.org/details/@uforbederlig 15Archive of Danish Literature, https://huggingface.co/datasets/danish-foundation-models/danish-dynaword/blob/ main/data/adl/adl.md. 7 Figure 4: Example pwa-scores. Quartiles. Aarhus Stifts-Tidende. Figure 5: Distribution of pwa-scores. 8 Usage notes Currently, we have made the data available in two forms. We ha...

arXiv 2010
[1819]

i” and “j

3 Methods Having trained some of the first neural models with Transkribus to deal with older forms of Danish handwriting, we saw an opportunity.8910 If emerging technologies could handle the high variation of handwriting, they might also be a match for old fraktur. In early 2022, we designed a prototype model based on about 100,000 words from eighteenth-c...

work page 2022
[1900]

In comparison, the second largest historical corpus of a similar quality from before 1900 is about 60 million words derived from novels 1870 to 1900.7 We estimate that the newspaper corpus eclipses all other Danish digitized text data of a similar quality from this period combined. The size as well as the diversity of the material – reflecting multiple co...

work page 1900

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

A World in Print: Introducing a Danish-Norwegian corpus of historical newspapers

A World in Print Introducing a Danish-Norwegian corpus of historical newspapers Johan Heinsen and Camilla Bøgeskov 2025-09-02 Abstract This Data Descriptor introduces the dataset Enevældens Nyheder Online (News during Absolutism Online). The Enevældens Nyheder Online (ENO) dataset provides a reconstruction of the contents of major newspapers in Denmark an...

work page Pith review arXiv 2025

[2] [2]

The disproportionate amount of data from the second half of the period is shown in figure 3, showing the distribution of individual texts pr. year. Figure 2: Coverage over time. From pages to texts One of our aims with the dataset was to explore certain genres of advertisements. These include advertisements relating to specific forms of consumption, adver...

work page 1989

[3] [5]

7 Figure 4: Example pwa-scores

14https://archive.org/details/@uforbederlig 15Archive of Danish Literature, https://huggingface.co/datasets/danish-foundation-models/danish-dynaword/blob/ main/data/adl/adl.md. 7 Figure 4: Example pwa-scores. Quartiles. Aarhus Stifts-Tidende. Figure 5: Distribution of pwa-scores. 8 Usage notes Currently, we have made the data available in two forms. We ha...

arXiv 2010

[4] [1819]

i” and “j

3 Methods Having trained some of the first neural models with Transkribus to deal with older forms of Danish handwriting, we saw an opportunity.8910 If emerging technologies could handle the high variation of handwriting, they might also be a match for old fraktur. In early 2022, we designed a prototype model based on about 100,000 words from eighteenth-c...

work page 2022

[5] [1900]

In comparison, the second largest historical corpus of a similar quality from before 1900 is about 60 million words derived from novels 1870 to 1900.7 We estimate that the newspaper corpus eclipses all other Danish digitized text data of a similar quality from this period combined. The size as well as the diversity of the material – reflecting multiple co...

work page 1900