Dating Undated Photographs and Documents
Which clues actually narrow the date of an undated photo or document, how AI can help gather them, and how to record a date range with its evidence in EDTF.
Read Article →Updates from the intersection of history and artificial intelligence.
Which clues actually narrow the date of an undated photo or document, how AI can help gather them, and how to record a date range with its evidence in EDTF.
Read Article →Small community archives hold decades of minutes, newsletters and photographs. Why description is the bottleneck, and how AI and volunteers can share the work.
Read Article →Most reels and tapes are described by a can label. How to build a usable film record with AI: time-coded shot lists, on-screen text, transcripts, human review.
Read Article →Keyword search quietly fails on OCR'd historical text. Why noise, old spelling and inflection defeat it, and the KWIC and recall checks that fix the gap.
Read Article →One ancestor can appear as Moshe, Movsha, Moszek and Morris. Why names split across scripts and borders, and how to search with Daitch-Mokotoff and AI.
Read Article →AI can draft descriptions fast, but a spreadsheet is not a catalogue. How to map AI output to Dublin Core, MODS or EAD, vocabularies, dates and provenance.
Read Article →Before an archive sends a single page to an AI system, it should know where that page is processed, what is kept, and who can see it. Five questions to ask any vendor, four deployment options, and how to tier a collection by sensitivity.
Read Article →Most OCR and handwriting-recognition failures are decided at the scanner, not in the model. Resolution, colour depth, file format, lighting and margins, and how to test them on 50 pages before you digitise 50,000.
Read Article →How documentary teams can use AI to log hundreds of hours of archival footage, transcribe interviews with timecodes, read on-screen text, and track rights clip by clip, and the questions of authenticity and ethics that no model can answer for you.
Read Article →What public Transkribus models (Dybbuk, Vaybertaytsh) actually cover in 2026, why pinkasim still fail off the shelf, and when a custom Yiddish HTR pipeline is the honest next step.
Read Article →SaaS archive platforms promise AI cataloging out of the box; custom pipelines promise accuracy on your material. A practical framework for deciding — including the hybrid route most institutions actually end up on.
Read Article →Most archives are sitting on decades of undescribed material. A backlog sprint — a scoped, time-boxed, AI-assisted cataloging push — is how small institutions turn 'someday' into a funded 8-12 week project.
Read Article →An AI answer without a source is an opinion. How provenance, audit trails, and page-level citation turn AI-assisted archival research into something a scholar, an editor, or a court can actually rely on.
Read Article →The people who remember are leaving us; the technology for listening to them has never been better. What AI genuinely changes in oral-history work — preparation, transcription, cross-referencing, writing — and what still depends entirely on the person holding the microphone.
Read Article →Modern Russian OCR quietly fails on anything printed before 1918: four abolished letters, final hard signs, worn newsprint. What actually works on the 1917 press — engines, fine-tuning, and the orthography decision every project must make.
Read Article →Photograph collections are the least catalogued material in most archives. AI vision can now find duplicates, read signage, place and date images, and draft records at scale — and there is a line it should not cross. A field guide to drawing that line.
Read Article →Not another generic tool the archive has to bend to. An AI cataloging engine that learns the domain, the fields, the language and the recording conventions of a specific archive — and produces finished records in the standard its existing system already accepts.
Read Article →A domain lexicon turns OCR guessing into informed decisions — how to build one that boosts Hebrew handwriting recognition, and when it backfires.
Read Article →An accuracy number without calibration is a bluff. How confidence scoring and the right threshold lifted Hebrew handwriting OCR from 56% to 90% precision.
Read Article →Three-layer confidence framework for cataloging Holocaust-era photographs with AI: visual, objects, persons — explicit ✓ ~ ? markers and when to defer.
Read Article →Which engine wins on historical Hebrew documents — rabbinical script, Rashi, Yiddish, Ladino? Accuracy numbers, fine-tuning workflow, and mistakes to avoid.
Read Article →Plain-language glossary for AI-assisted historical research: OCR, HTR, RAG, embeddings, knowledge graphs, NER, pinkasim, yizkor books, and 18 more terms.
Read Article →
A structured database consolidating dozens of source types — press, diplomacy, diaries, minutes — from one decisive year into a searchable knowledge base.
Read Article →How archival research combined with AI turned a hundreds-of-pages memoir into a scholarly edition with hundreds of footnotes — in months, not years.
Read Article →How we used OCR and AI agents to turn thousands of pages of historical newspapers into a searchable database — and surfaced long-forgotten quotes.
Read Article →How we used LLMs and research agents to turn dozens of family memoirs into a unified historical narrative — verifying facts and revealing hidden links.
Read Article →How AI research agents working in parallel across dozens of sources reconstructed a historical event nearly erased from the record — a case study.
Read Article →How artificial intelligence is helping researchers uncover, cross-reference, and preserve Holocaust testimonies and documents at unprecedented scale.
Read Article →Traditional OCR cannot read handwritten documents. Modern HTR technology, powered by deep learning, is finally making centuries of handwritten records searchable.
Read Article →How AI is transforming archival work — from passive image scans to extracting meaning. Three layers of an AI-native archive, with a concrete worked example.
Read Article →Autonomous AI research agents can now conduct preliminary historical research across multiple archives simultaneously. Here's how they're changing academic work.
Read Article →Knowledge graphs connect isolated documents into networks of people, places and events — revealing patterns invisible to traditional research.
Read Article →How researchers query thousands of historical documents in plain English — with source-cited answers. Setup, and where RAG still fails.
Read Article →Museums, archives, and libraries face an urgent race against time. AI-powered digital preservation offers new tools to protect cultural heritage for future generations.
Read Article →Discover how artificial intelligence is revolutionizing family history research, from deciphering old letters to connecting scattered records across continents.
Read Article →Off-the-shelf OCR fails on historical Hebrew. Here's the AI workflow archivists use to recover RTL manuscripts, Rashi script and worn pages.
Read Article →