Sebastian Majstorovic

Made some improvements to the OCR scripts onboarding in my uv-scripts collection. First run is now a small example: seven scanned NASA pages in, Markdown out, one HF jobs command. It also works on your own images or a folder of PDFs. You can pick from 28 OCR models. huggingface.co/datasets/uv-...

Need a historical illustration? Search 1.49 million images from British Library books and Britannica (1500s–1920s). Find similar images, download full-resolution images or transparent PNG cutouts, and explore the sources. huggingface.co/spaces/davan...

OCR for Japanese manga, Swedish handwriting or Arabic print? There’s a growing range of OCR models on @hf.co: VLMs, dedicated text recognisers and complete OCR pipelines. I’ve gathered 41 models into four collections, with short notes to help you choose: huggingface.co/collections/...

A 15.9M-parameter specialist OCR model outperforms many VLMs hundreds of times larger. On a 2,165-page historical eval dataset, Kraken PP-OCRv6 ranks #4 on reading CER and #1 when long-s, ligatures and case are preserved. huggingface.co/spaces/fineb...

Every illustration in the British Library's 19th-century book images dataset now has an instance mask: 1.02M clean cutouts of public-domain engravings, no image generation involved! One small segmentation model, $3.24 of compute. Model, masks and a search Space with cutout view all open.

Interested in participating in the 2027 Central European History Convention July 15-17, 2027? Submit a paper proposal! Deadline September 13, 2026. Check here for all the different options of how to apply to be part of the program! ceh-c.univie.ac.at/news/call-fo... Please share far and wide!

FineBooks wants to help re-OCR the public domain to unlock a rich source of data for AI, for researchers, and for enjoyment. Step one: work out which modern OCR models are actually good enough.

Are open OCR models good enough to unlock historical knowledge? @danielvanstrien.bsky.social and I built a leaderboard using six expert-transcribed volumes from the @biodivlibrary.bsky.social sky.social. Meet FineBooks, from @hf.co and @eleutherai.bsky.social! huggingface.co/blog/fineboo...

Uploaded a dataset of 1,080,814 public domain images, mostly from 19th-century books, to the Hugging Face Hub: huggingface.co/datasets/big... You can also do semantic search against the images here: huggingface.co/spaces/davan...

Heon Cha Haus was a great tip for Seoul @tedunderwood.com. I had their signature iced tea and an extremely tasty Red Bean & Cream Rice Cake. Not to mention the wonderful service! #DH2026

Enjoying the weekend in Seoul before #DH2026 kicks off next week in Daejeon @dh2026daejeon.bsky.social.

I saw this and had to share.

We are so very grateful for this honor! Thank you to our supporters and everyone who voted. Congratulations to everyone who was nominated -- we feel very privileged to be counted among your incredible work.

Congratulations to the 2026 Data Integrity Award winners - @datarescueproject.org - Gina Plata-Nino of @fracposts.bsky.social - @mapresearch.bsky.social & Williams Institute - @hudgov.bsky.social & FEMA - Disaster Recovery Data Portal Details in 🧵

Read the feasibility study for the European Books Data Commons and discover the next steps! Explore how European libraries can make the full text of millions of digitised books available for research, innovation and contribute to a European AI infrastructure built on public values: bit.ly/4gs09tY

#DulaPeepShow #DuaLipa

@storytracer.com shows how volunteers or #GLAMS workers can be responsible actors in the current political climate by protecting datasets from war, political censorship and defunding! #glamlabsfutures

I think VLM-based OCR might finally be close to working on historic newspapers! Many models I've tried before failed i.e. hallucinations, repetition loops, context overflow. Surya OCR 2 (a 650M model!) does a very good job!

Ran the same 650M model across 7 languages: German, Polish, Finnish, Russian, Serbian, Estonian, Swedish via @europeana.bsky.social newspapers. Still some challenging layouts, but pretty mind-blowing to me that a 650M model is doing this when a year ago 72B VLMs failed very often.

Read our newly published paper on AI 'The case for Public AI: making it happen with cultural heritage.' 400+ professionals, one shared position - cultural heritage should shape AI, not just feed it. 👉 bit.ly/4xLdyU3 #PublicAI #AI #ArtificialIntelligence #CulturalHeritage #DataSpace

Unfortunate to see this outcome but the fight for the panels at Washington’s house will continue in the community! This is our history and we want to #SaveOurSigns

In the 90 minutes after @theguardian.com article went live, BHL received more than US$2,500 in donations. Thank you to everyone helping keep biodiversity knowledge free and open. Please read, share, and help us build momentum.

A federal judge in Massachusetts has ordered the Trump administration to restore all signs that were changed or removed at national parks across the country as part of President Donald Trump’s controversial directive last year.

OCR for Ancient Greek is constrained by a lack of open training data. But the deeper challenge tackled by 2 papers from Inria is doing OCR and document structure recovery together: section hierarchies, milestone numbering, marginal references. Paper 1: arxiv.org/html/2603.02...

You can now run SQL over 2.19 BILLION web pages — zero download. @commoncrawl.bsky.social April 2026 crawl + URL index are on Hugging Face Storage Buckets. DuckDB reads it straight over hf:// — I counted all 2.19B in ~35s. Or point your own agent at it 👇 huggingface.co/spaces/davan...

NuExtract3 (4B, Apache-2.0) does OCR *and* structured extraction. Point it at a dataset of scanned index cards + a JSON schema → clean catalog JSON One command on huggingface Jobs. (or skip the schema for plain Markdown OCR) Script + dataset 👇

We're announcing two changes CourtListener API access: 1. Full API access is now open to everyone, including the PACER APIs that previously required a conversation with us. 2. Higher tiers are available through FLP memberships (including edu!) or commercial agreements. 👇 free.law/2026/05/07/a...

The Guardian: “‘Things Were Going Dark Left and Right’: The Race to Save US Government #Datasets Before They’re Deleted” www.infodocket.com/2026/05/07/t... @datarescueproject.org @envirodgi.bsky.social @archive.org #data #libraries #librarians

Thanks to @theguardian.com for highlighting the people involved in the effort to save and demonstrate the importance of federal public data. ❤️🛟 www.theguardian.com/us-news/2026...

As we head into the weekend, and near the end of our #BobsBurgers month of memes, I wanted to share this joyous depiction of our truly amazing DRP volunteers. They give me hope without intending to; they're creating the good they want to see in the world; they are all gold stars.

“[T]he popularity of AskHistorians demonstrates that there can and should be more history-related jobs in both academic and public-facing settings. People want history to understand the world, and I can also tell you that those people do not want the AI-generated version of their answers.”

We’ve signed the #OpenHeritageStatement to support a global call for equitable access to #PublicDomain heritage in the digital environment! Discover more about the Statement, how it links to our advocacy for open #CulturalHeritage and how you can sign ➡️ bit.ly/4sB1Ur0 @creativecommons.bsky.social

After Trump took office last year, the US Holocaust Museum quietly removed educational material about American racism from its website and canceled a workshop on the "fragility of democracy," @iriesentner.bsky.social reports. www.politico.com/news/2026/04...

You never know what data will be used for! I uploaded a @britishlibrary.bsky.social dataset to Hugging Face in 2022. IIRC one of my first PR to a HF repo! 4 years later, someone trains a Victorian chatbot on it More libraries should be sharing their public domain collections for AI to build on!

📣 New historical data visualization! "How Fast Was the Mail?" is an interactive map showing how long information took to travel across the US between 1882-1908: cblevins.github.io/mail-time/ +

I’m at #HSP2026 at MIT this week! I’ll be giving a talk Friday at 5:25pm entitled “Structural Priming Effects in Language Models are Less Human-like in Languages Other Than English”. Looking forward to chatting to everyone!

Interesting new piece from @iatp.bsky.social showing how confidence in USDA data is eroding under the Trump Administration, with serious consequences. www.iatp.org/farming-risk...

New DRP post: #Philly interpretive panels returned to the President's House. We congratulate the local community that fought HARD for their return. We hope this inspires other communities to #SaveOurSigns www.datarescueproject.org/independence...

Small update to #tinyIIIF, my no-nonsense #IIIF server! You can now choose your image server during setup: • IIPImage • Cantaloupe Running small- to mid-sized collections? Teaching with IIIF materials? Building IIIF-enabled tools? Check out tiny.iiif! github.com/rsimon/tiny-... #DigitalFriday

Re-OCR'd the complete 1771 Encyclopaedia Britannica (2,724 pages) with a single command on @hf.co Jobs. - 0.9B model (GLM-OCR) ~$0.002/page ~$5 total on an L4 GPU Before (old Tesseract ocr) → After

Get ready for live updates and quotes from @lyndamk.bsky.social and @mikalarae.bsky.social's #IDCC26 Keynote!

Yay! The first image served from my new #tinyIIIF test instance. I'm running it on a 2 CPU/4GB RAM VM. Seems to be the absolute minimum & performance is pretty slow. If anyone has advice on which specs I'd need to get more #IIIF speed out of Cantaloupe – let me know!

Announcing our latest paper: CommonLID In collaboration with @commoncrawl.bsky.social @mlcommons.org @jhu.edu we built a LID benchmark on actual Common Crawl text covering 109 languages. Existing evaluations overestimate how well LangID works on web data. arxiv.org/abs/2601.18026

It's official. One year ago today we formalized as the Data Rescue Project. We know this work is exhausting and are so grateful to everyone who has showed up, day in and day out.

🎉Congrats to the winners of the 2025 RDAP Work of the Year Award, The Data Rescue Project! (@datarescueproject.org)🛟🏆 This award acknowledges the work’s impact on the wider research and scholarly communication ecosystem in support of RDAP’s mission and values. rdapassociation.org/news/13593532

Built a 2.5MB image classifier that runs in the browser in an evening with Claude Code. I used a dataset I labelled in 2022 and left on @hf.co for 3 years 😬. It finds illustrated pages in historical books. No server. No GPU.

Very happy to introduce a new tool, BookReconciler! You can take spreadsheets with book data and add subject headings, descriptions, ISBNs, HathiTrust IDs, & more. You can also cluster editions & variations of the same "Work." Led by @thisismattmiller.com and supported by @post45data.bsky.social.