PreviewNot yet reviewed. Cards on this preview are seed data waiting for the editors’ approval, so check anything before you cite it.
The Digital Historian’s Toolbox

Preview · 19 tools · 6 datasets · 75 projects · seeded October 2026

The tools and datasets historians actually used, and where.

Pull out a drawer to see what’s filed inside. Every card separates research that used a tool from the benchmarks that tested it.

  1. Mistral OCR (2 uses)
  2. World Historical Gazetteer (1 use)
  3. Gephi (1 use)
  4. Voyant Tools (1 use)
  5. eScriptorium (1 use)

Research + corpus-building uses. Benchmarks don’t count.

The drawers

A tool can be filed in more than one drawer
  • DRAWER I

    AI & History

    Machine learning, vision-language models and speech models. Every card here carries notes on where your sources go, how it was tested, and how it fails.

    Guide cards

    Counts research and corpus-building uses in the window, plus “I’m using this now” check-ins. Benchmarks don’t count.

    • tool · No. 019 AI

      PaddleOCR-VL

      Small, free open-weight model from Baidu that reads scans into structured text, tables and formulas in 109 languages. Runs on your own computer.

      At a glance
      Maker
      Baidu (PaddlePaddle)
      Released
      Oct 2025 (v1.6: May 2026)
      Licence
      Open source (Apache-2.0)
      Cost
      Free to run locally
      Runs
      Your own machine; offline Docker image
      Skill
      Programming
      Historian’s notes
      Where your sources go Local use keeps documents on your hardware. Hosted service terms unverified.
      How it’s been tested Scores 80 overall on olmOCR-Bench but only 37.8 on its old-scans set. Its own benchmarks don’t test historical documents.
      Known problems on historical sources A researcher digitising 20 million census pages rejected it after fine-tuning produced stray Chinese characters. Skipping the layout step can add invented text.
      Used in research0

      No published research uses found yet. Using it in a project? Tell us.

      Used to build corpora1
      Tested in1
      • socOCRbench · Noah Dasanaike (Harvard, Government), 2026
      Visit website
      Add a project that used it See it on the network Suggest a correction

      Nobody has said they’re using it in 2026 yet. “I’m using this now” asks you to sign in with ORCID. One count per person per year; it adds to Trending, and names are never shown.

      Last updated 14 Oct 2026 Not yet reviewed by the editors · Version history