Cultural preservation: AI for historical documents and oral histories

Every place we work has a human history. San Domingo Arbory and Woodland Hilltop Arbory are in the homeland of the Miwok people, who still live in the foothills, and heritage information will be handled with tribal consultation. The written record we have gathered begins in the Gold Rush years and runs through the homesteads, ranches and mining claims that followed. This project gathers that record, reads it carefully, and shares it one source or topic at a time.

The sources

The archive behind the site histories is private and filed by site, and a catalog records what each item is, its date and what it covers. The catalog lists more than 350 items: documents, newspaper articles, maps, books and photographs.

• San Domingo Arbory. Federal homestead patents of 1888 to the Ponte family and their Briseno neighbors along San Domingo Creek, water-right notices from the 1880s and 1890s, deeds and assessment rolls from 1887 to 1960, an 1871 federal land survey, and newspaper accounts and local histories. Together they trace the original Ponte and Valente homesteads that make up the arbory. Federal homestead case files for the Ponte and Briseno homesteads and a neighbor’s, which should hold the claimants’ sworn testimony, are image-only scans still to be transcribed.
• The Briseno and Ponte families. Census pages from 1880 to 1930 and voter registers from the 1890s and 1900s, a marriage record, anniversary and obituary notices, a family genealogy, and a reminiscence of ranch life in the 1920s by a grandson born on the ranch, published in a local historical society bulletin in 2014. Many of the census pages are too faint for text recognition and need transcribing by hand. Genealogy that lists living relatives stays private.
• Mining records. Claim notices filed by members of the Ponte family between 1878 and 1922, and the patent papers for a mining company’s placer claim, entered in 1900. They show the family filing mining claims for more than four decades, alongside the ranch. We describe the records, never where the workings are.
• Woodland Hilltop Arbory. Early twentieth-century mining-claim notices, a federal mineral patent and a deed for two quartz claims that appear to cover the hilltop, alongside the records of a short-lived mining and smelting promotion at Murphys. We will publish the hilltop’s history from them once a person has checked them against the originals.
• Around Murphys. Newspapers from the 1850s onward, mining reports and published local histories that cover more than one site.

The record has gaps, and we say so. No document we have read yet covers most of Woodland Hilltop’s twentieth century, and dozens of newspaper clippings are image-only scans still to be read.

How the work is done

• Transcription. AI tools, Claude among them, read scans of handwriting and old typescript and draft transcriptions as written: original spelling kept, unclear words marked, nothing silently corrected.
• Summaries with citations. Draft summaries and timelines cite a source and page for every claim, keep what a source says apart from what we infer, and list where sources disagree.
• People verify. A person checks every transcription line by line, and every name, date and quote against the original, before anything is published. AI can be fluent and confidently wrong, and file names can mislead too: one 1905 document in the Woodland Hilltop files, filed under a name that read like a sale, turned out on reading to be a mortgage securing the mining company’s bonds.
• Oral histories, with consent. What people who know this land tell us is used only as their written consent allows: how a recording may be transcribed (including with AI), quoted and named. Narrators review their words before anything is published, and can withdraw.
• Disclosure. Every history piece made with AI help says so at the end.

What we don’t publish

• No locations of archaeological or cultural features, and no maps, coordinates or legal descriptions from the records. Old mine workings and homestead sites are described, never located.
• Nothing about Miwok heritage beyond this acknowledgment until it has been through tribal consultation, with site-location confidentiality.
• No private individuals who are still living, even when they appear in public records, unless they agree to be named.

The history section of the San Domingo Arbory page draws on this work, and history posts on the Research Blog will add to it one source at a time.

Get involved

If you hold family records, photographs or memories of these places and would be willing to share them, would like to be interviewed, or see something we should correct, please contact us.