Jacob Polay

Writing & Talks

Publications, public scholarship, and presentations.

Publications

  1. Clifford, Jim, Jacob Polay, Jessica Jack, Mark Humphries, and Lianne C. Leddy. "Reading the Archive by Machine: An OCR Benchmark for Historians, 1612–1921." Working Papers in Critical Search, no. 4 (2026). https://working-papers-in-critical-search.github.io/paper-004-ocr-benchmark/.

    Cite

    Clifford, Jim, Jacob Polay, Jessica Jack, Mark Humphries, and Lianne C. Leddy. "Reading the Archive by Machine: An OCR Benchmark for Historians, 1612–1921." Working Papers in Critical Search, no. 4 (2026). https://working-papers-in-critical-search.github.io/paper-004-ocr-benchmark/.

    BibTeX

  2. Polay, Jacob. "Circumventing the Garbler." Intersections 8, no. 1 (2025): 28.

    Cite

    Polay, Jacob. "Circumventing the Garbler." Intersections 8, no. 1 (2025): 28.

    BibTeX

Public scholarship

Most of these posts belong to AI and Collaboration, a series on Active History initiated by Chloë Farr, Jessica Jack and me.

  1. Jack, Jessica, Chloë Farr, and Jacob Polay. "Not Just Farming: How an LLM Helped Me Study Saskatchewan’s Smallest Settlements." Active History, October 14, 2026. https://activehistory.ca/blog/2026/10/14/llm-saskatchewans-settlement-history/.

    AI and Collaboration series. Lead author Jessica Jack, with Chloë Farr and Jacob Polay. How a language model helped join datasets built for narrower questions into a typology of Saskatchewan's smallest settlements.

    Cite

    Jack, Jessica, Chloë Farr, and Jacob Polay. "Not Just Farming: How an LLM Helped Me Study Saskatchewan’s Smallest Settlements." Active History, October 14, 2026. https://activehistory.ca/blog/2026/10/14/llm-saskatchewans-settlement-history/.

    BibTeX

  2. Polay, Jacob, Chloë Farr, and Jessica Jack. "Can You Read Two Million Pages? An LLM Can Help." Active History, October 7, 2026. https://activehistory.ca/blog/2026/10/07/named-entity-recognition-large-language-models/.

    AI and Collaboration series. Lead author Jacob Polay, with Chloë Farr and Jessica Jack. How a knowledge graph and GraphRAG, built from 20,000 early modern documents, can help historians read a corpus far too large to read page by page.

    Cite

    Polay, Jacob, Chloë Farr, and Jessica Jack. "Can You Read Two Million Pages? An LLM Can Help." Active History, October 7, 2026. https://activehistory.ca/blog/2026/10/07/named-entity-recognition-large-language-models/.

    BibTeX

  3. Polay, Jacob, Chloë Farr, and Jessica Jack. "Stop Choosing Tools. Start Building Them." Active History, September 30, 2026. https://activehistory.ca/blog/2026/09/30/building-ai-tools-historical-research/.

    AI and Collaboration series. Lead author Jacob Polay, with Chloë Farr and Jessica Jack. Why historians should build their own tools instead of choosing off-the-shelf ones, using named-entity recognition for early modern documents as the example.

    Cite

    Polay, Jacob, Chloë Farr, and Jessica Jack. "Stop Choosing Tools. Start Building Them." Active History, September 30, 2026. https://activehistory.ca/blog/2026/09/30/building-ai-tools-historical-research/.

    BibTeX

  4. Farr, Chloë, Jessica Jack, and Jacob Polay. "The AI-Enabled Boom in Document Transcription." Active History, September 23, 2026. https://activehistory.ca/blog/2026/09/23/ai-enabled-document-transcription/.

    AI and Collaboration series. Lead author Chloë Farr, with Jessica Jack and Jacob Polay. On the new wave of AI transcription tools for printed and handwritten sources, and why outputs built to train language models often fall short of what historians need.

    Cite

    Farr, Chloë, Jessica Jack, and Jacob Polay. "The AI-Enabled Boom in Document Transcription." Active History, September 23, 2026. https://activehistory.ca/blog/2026/09/23/ai-enabled-document-transcription/.

    BibTeX

  5. Polay, Jacob, Chloë Farr, and Jessica Jack. "The Archive is Not a Toy: The Hidden Problems of a 'Vintage' AI." Active History, July 13, 2026. https://activehistory.ca/blog/2026/07/13/the-archive-is-not-a-toy-the-hidden-problems-of-a-vintage-ai/.

    A critique of Talkie-1930, a "vintage" language model trained on pre-1930 text. It argues that the model aestheticizes a narrow, elite, colonial archival record, presents a legally motivated corpus cutoff as historically meaningful, and deploys a research tool as a public chatbot without adequate framing.

    Cite

    Polay, Jacob, Chloë Farr, and Jessica Jack. "The Archive is Not a Toy: The Hidden Problems of a 'Vintage' AI." Active History, July 13, 2026. https://activehistory.ca/blog/2026/07/13/the-archive-is-not-a-toy-the-hidden-problems-of-a-vintage-ai/.

    BibTeX

Talks and presentations

  1. “Fitting Machine to Page.” With Keith Pemberton. AI and History Conference, American Historical Association, Baltimore.

    An invited paper on a panel on AI methods, about OCR and its consequences for historical practice.

  2. “Reading the Unreadable: Optical Character Recognition of Early Modern Print.” Canadian Society for Digital Humanities, Montreal.

    Two chapters of my MA thesis, presented to the largest gathering of digital humanists in Canada.

  3. “Solving OCR: Using olmOCR to Follow Commodities across the British World.” With Jim Clifford. Environmental Digital Humanities Seminar (Lancaster and Manchester), Online.