IDI (@institutional.org)

institutional.orgExtending Institutional Books with enriched text and 22M extracted images | IDIThe Institutional Data Initiative at the Harvard Law School Library expands the ~1M-book Institutional Books dataset with paragraph-level text analysis and 22M extracted images, enabling new forms of ...

8 Sept

Institutional Books is expanding. The Institutional Data Initiative @institutional.org at @harvardlawadmin.bsky.social Library has added paragraph-level text analysis and 22M extracted images to our work on ∼1M books, opening new possibilities for research, experimentation, and discovery. 🧵

Together, these projects establish Institutional Books as an ongoing resource that can grow through continued research and community use. The open-source pipelines can be adapted to new collections, models, languages, and questions: institutional.org/posts/extend...

8 Sept

The Institutional Data Initiative at Harvard @institutional.org works with knowledge institutions to increase the availability, diversity, and responsible use of training data for AI. Reach out and join us.

Enriched Text contains 217B tokens from 983K @harvardlibrary.bsky.social volumes, organized into 1.39B annotated paragraphs across 250 languages. Researchers can use the dataset to analyze language, repeated passages, line breaks, and paragraph predictability.

8 Sept

Visual Elements extends the dataset to images embedded in scanned pages, producing 22.6M visual elements – including illustrations, photographs, charts, musical notation, and decorative materials – linked to their source volumes and pages.

institutional.orgMaking 1.47 Million Historical Newspaper Pages Available as High-Quality, Granular Data | IDIThe Institutional Data Initiative at the Harvard Law School Library and Boston Public Library release an open dataset and processing pipeline that make historical newspapers more accessible for resear...

2 Sept

When a newspaper becomes a wall of OCR text, columns collapse, headlines mix with ads, and reading order disappears. The Institutional Data Initiative (@institutional.org) and Boston Public Library (@bpl.boston.gov) built a pipeline and dataset to make historical newspapers more accessible.🧵

The dataset supports more than keyword search: you can follow reading order, identify people and places, and examine how newspaper design and advertising changed across time. Institutional Newspapers includes the dataset, models, and an open-source processing pipeline:

2 Sept

The Institutional Data Initiative at the @hls.harvard.edu Library works with knowledge institutions to increase the availability, diversity, and responsible use of training data for AI. Reach out and join us.

Institutional Newspapers contains 1.47 million pages from the Boston Public Library’s collections, representing 16.3 billion tokens and a multilingual record of Boston’s communities across 73 languages.

2 Sept

Instead of treating each page as one image, our pipeline divides it into individual crops. Those crops are processed as articles, advertisements, photographs, cartoons, headings, mastheads, or empty space, with OCR tailored to each segment.

IDI reposted

institutional.orgInstitutional Proceedings: 186 years of university decisions. | IDIIDI and the University of Michigan Library have collaborated to release two centuries of the university’s governance decisions as a structured dataset that can be searched, analyzed, and reused.

13 Aug

AI enables new forms of research, but only where information is computationally accessible. Today, @institutional.org and @umich.edu Library are releasing a structured dataset that makes 186 years of U-M’s Board of Regents Proceedings newly accessible. institutional.org/posts/instit...

IDI reposted

arxiv.orgGRIN Transfer: A production-ready tool for libraries to retrieve digital copies from Google BooksPublicly launched in 2004, the Google Books project has scanned tens of millions of items in partnership with libraries around the world. As part of this project, Google created the Google Return Inte...

20 Nov 2025

Replying to @leppert.me

Even if you're not a partner library, you might be curious about what it's like to work with GRIN. Our technical report has a wealth of details. arxiv.org/abs/2511.11447

IDI reposted

institutional.orgInstitutional Books | Institutional Data InitiativeInstitutional Books 1.0 is our first release of public domain books. This set was originally digitized through Harvard Library’s participation in the Google Books project..

20 Nov 2025

Replying to @leppert.me

We're also sharing the pipeline we developed for Institutional Books that seamlessly dedupes, classifies, and enhances the data once GRIN Transfer brings it down. www.institutional.org/tools

IDI reposted

institutional.orgAnnouncing the release of GRIN TransferGRIN Transfer, an open source tool that allows Google Books partner libraries to more easily access their Google Books collection.

20 Nov 2025

When libraries join Google Books, Google not only scans their books, it also makes a wealth of image, OCR, & metadata available to them via the Google Return Interface (GRIN). But working with GRIN can be challenging, so we're releasing a tool to make it easier. www.institutional.org/posts/grin-t...

tinyurl.comWelcome! You are invited to join a meeting: IDI Talk with Michele Dolfi & Peter Staar on SmolDocling. After registering, you will receive a confirmation email about joining the meeting.Please join us for a talk with Michele Dolfi & Peter Staar from IBM Research in Zurich to discuss their work on SmolDocling, an “ultra-compact” model for diverse OCR tasks.

18 Sept 2025

Join us tomorrow at 10AM EST: tinyurl.com/y3ye6cz6

IDI@institutional.org

Can a small visual language model read documents as effectively as models 27 times its size? Next Friday, IDI will host Michele Dolfi and Peter Staar from IBM Research Zurich to discuss their work on SmolDocling, an “ultra-compact” model for diverse OCR tasks.

harvard.zoom.usWelcome! You are invited to join a meeting: IDI Talk with Michele Dolfi & Peter Staar on SmolDocling. After registering, you will receive a confirmation email about joining the meeting.Please join us for a talk with Michele Dolfi & Peter Staar from IBM Research in Zurich to discuss their work on SmolDocling, an “ultra-compact” model for diverse OCR tasks.

IDI reposted

20 June 2025

This Monday, @institutionaldatainitiative.org will host Petr Knoth to share his experience leading CORE ("The world’s largest collection of open access research papers") as the rise of AI brings new meaning, and challenges, to stewarding knowledge repositories. Join us virtually via the link below.

IDI reposted

harvard.zoom.usWelcome! You are invited to join a webinar: Open AI Development. After registering, you will receive a confirmation email about joining the webinar.For AI to truly benefit society, it must be built on foundations of transparency, fairness, and accountability—starting with the most foundational building block that powers it: data. Not long ago, ...

16 June 2025

Replying to @leppert.me

Cohosted by @institutionaldatainitiative.org and The Berkman Klein Center. harvard.zoom.us/webinar/regi...

huggingface.coinstitutional/institutional-books-1.0 · Datasets at Hugging FaceWe’re on a journey to advance and democratize artificial intelligence through open source and open science.

Today we released Institutional Books 1.0, a 242B token dataset from Harvard Library's collections, refined for accuracy and usability. 🧵

We look forward to growing Institutional Books through community. We welcome collaboration from researchers and model makers as we: - Evaluate the dataset’s impact on model outputs - Continuing to refine our OCR pipelines View the dataset on Hugging Face: huggingface.co/datasets/ins...

institutionaldatainitiative.orgInstitutional Books | Institutional Data InitiativeInstitutional Books 1.0 is our first release of public domain books. This set was originally digitized through Harvard Library’s participation in the Google Books project..

We hope Institutional Books will be the beginning of a process that makes millions more books accessible to the public for a variety of uses. We welcome feedback as we continue to expand this dataset, refine its contents, and sharpen our process. www.institutionaldatainitiative.org/institutiona...

Today we released Institutional Books 1.0, a 242B token dataset from Harvard Library's collections, refined for accuracy and usability. 🧵

We included extensive volume-level metadata with both original and generated components, such as results from text-level language detection.

As part of our refinement work, we supplemented the original OCR-extracted text with a post-processed version that utilizes line detection to reassemble the text according to the line type.

Today we released Institutional Books 1.0, a 242B token dataset from Harvard Library's collections, refined for accuracy and usability. 🧵

We analyzed the dataset’s coverage across time, topic, and language and found: - 40% of English text + long tail of 254 languages - 20 clear topical tranches - Largely published in the 19th and 20th centuries Technical report here: arxiv.org/abs/2506.08300

IDI reposted

14 Apr 2025

The @institutionaldatainitiative.org is proud to support The New Commons challenge. $100k grants along with mentorship. Let's get impactful data into the AI ecosystem.

The GovLab@thegovlab.org

(1/4) CALL FOR APPLICATIONS FOR DATA COMMONS FOR AI 🏆Today, The Open Data Policy Lab (a collaboration btwn The GovLab & @microsoft.com launched The New Commons Challenge—an innovation challenge to foster the creation of data commons that can support generative AI developed in the public interest.

IDI reposted

institutionaldatainitiative.orgUsing AI to Accelerate Digitization at Boston Public LibrarysToday, as part of our mission expansion, we’re announcing a collaboration with BPL to develop AI-driven tools capable of accelerating new digitization of large collections at libraries across the worl...

12 Mar 2025

As the @institutionaldatainitiative.org expands its mission, we’re announcing a collaboration with @bpl.boston.gov to develop AI-driven tools capable of accelerating new digitization at libraries across the world, starting at the Boston Public Library. institutionaldatainitiative.org/posts/using-...

IDI reposted

institutionaldatainitiative.orgExpanding Our Mission: An Open Call for CollaboratorsToday, we’re pleased to announce an open call for institutional collaborators as new support expands the research capacity of the Institutional Data Initiative.

5 Mar 2025

I'm pleased to announce we're expanding our mission at the @institutionaldatainitiative.org with an open call for institutional collaborators, new digitization at Harvard Law School Library, and additional support to advance this work. institutionaldatainitiative.org/posts/open-c...