The Story Behind the Incredible Resource archival.name

If you've been anywhere near the Ukrainian genealogy community in the last few weeks, you've heard about archival.name (which I wrote about here earlier this week).  I virtually sat down with its creator, Luke Rothman, to find out where it came from, how it works, and what comes next.  I'm a huge fan of Luke's work; a record I found on the site last night made a 99-year-old Holocaust survivor cry.

 

The marriage record of my cousin, a 99-year-old Holocaust survivor

 

The origin story of this site starts with a missing letter. Someone asked Luke for research help, and in the course of that work he found that the Krakow 1921 census has a handwritten index missing the letter S. The census runs about 20,000 pages, and the index covers about 3,000 of them. He wondered whether Claude could use the index as a guide to find the missing entries.  It has only been about a month from when he (successfully) tried this, and the site has been up for two weeks--then it took less than two weeks to do similar work with different content, and it took only two days from to build and launch the site, which is pretty incredible.

I tried something similar about six months ago and gave up. The AI guessed at names, and correcting its errors took longer than reading the pages myself. Luke had the same experience in reverse. He ran tests on models from the past year and found that attempting this just three or four months earlier would have cost roughly four times as much and been far less accurate. A newer, cheaper Gemini Flash model that arrived in early September, just before he began in late September, made the economics workable. He's candid that the timing was lucky--and we are lucky recipients of this work.

Luke focused on material that isn't indexed anywhere, partly to avoid duplicating other people's work. At this point, the site includes:

Lwów/Lviv, beyond the vital records:

• Deaths records from1864, marriages from 1872, house lists from 1856–1889, and conversions run from the 1840s to the 1940s. Births cover 1900–1924. 
• Soviet civil records 1940–1946 (305,835 entries)
• Jewish name lists 1937–1943
• Jewish school records 1920–1938
• Emigration Office applications, Palestine Office emigrants and Hechaluc members, 1920–1939
• Conversions, 1840s–1940s

Kraków:

• Jewish registers 1827–1941 (186,000 entries)
• Kraków and Podgórze censuses 1857–1921 (over 700,000 entries)
• House lists

Galicia:

• Jewish vital records: 1848–1942, over 800,000 entries, including the TsDIAL books for Lwów and Eastern Galicia
• Censuses: Nowy Sącz, Rzeszów, Łańcut 1910 and Tarnopol 1910
• Stanisławów: conscript registers, residents 1920–1939, Soviet civil records 1940–1945 and war-damage claims
• War and its aftermath: Polish repatriation lists 1944–1946, Jewish Social Self-Help questionnaires 1940–1941, and Galician and Bukovinian war refugees in Vienna 1914–1915

Bukovina: Czernowitz and Sadagora Jewish registers 1871–1930.

Central Poland:

• Jewish registers 1915–1942 (211,000 entries)
• Community papers and lists 1918–1942

Ukraine:

• Jewish metrical books of the Russian Empire 1863–1922, from the Kyiv, Volhynia, Podolia, Poltava and Yekaterinoslav provinces
• Starokostiantyniv civil registration 1921–1930
• Kharkiv census of December 1941
• Kharkiv civil registry indexes 1920–1941
• Volhynia Jewish registers 1920–1938
• Łuck and Równe community lists 1919–1939

Vienna: Lehmann’s address books 1915–1942 (2.4 million people).

Soldiers:

• Austro-Hungarian casualty lists 1914–1918 (2.6 million entries, every faith and nation)
• Tarnów regiment recruits 1914–1915 

Census and vital records from regions with little or nothing online.

Phone books and directories from Eastern and Central Europe. They are cheap to process and, once compiled, give the system many correlated names and towns to improve its reading of handwritten material.

Archive finding aids from Ukraine, Belarus, Latvia, Lithuania, Moldova and elsewhere, translated and searchable. Anyone who has downloaded these over a flaky connection and tried to OCR them (like me!) will appreciate this one.

Galicia and Bukovina village reference drawn from a 1900 Austro-Hungarian census volume.

 

Behind the scenes, Luke is building a geotagged database of several hundred thousand place names in many languages, alphabets and time periods. It helps the system decide which "Kolomyia" a record means, and he is working to improve it. He is also linking records through the cross-references written on 20th-century documents, such as a marriage record that points to the groom's birth record. This could be a long-term game changer.

Because the transcriptions capture every word on the page, not just names, parents and professions, I'd say they are in some ways richer than traditional indexes.

Advanced researchers will want to know the limits. According to Luke:

• Cyrillic handwriting before about 1880 is beyond what any AI does well right now--which also explains why many of the included records are from the early twentieth century.
• Handwritten Hebrew and Yiddish are the weakest areas, and printed Hebrew suffers on poor scans.
• Printed Cyrillic and tidy modern handwriting are generally fine.

He wouldn't be surprised if even the older Cyrillic hands become feasible by early 2027.  And with the speed of AI improvements, I wouldn't doubt that would be the case.

Beyond the weakness of the models on earlier Cyrillic handwriting, Luke focuses on 20th-century records on purpose. Many people just starting out don't know their grandparents' names or hometowns, and the records that would tell them aren't indexed. Once they find those, they may be drawn to the older records that volunteers have painstakingly transcribed. He sees his site as a bridge, not a replacement.

A lot of people have asked me whether they can trust a new site run by someone they don't know. Luke and I know some people in common, so I'm personally okay on that front.  But I heard concern from those who are wondering why you have to log in to use the site.  Per Luke, he is doing that to let individuals use the data freely while preventing bulk scraping, so the whole dataset doesn't end up behind a commercial paywall.  He doesn't share or sell emails, and you won't receive emails unless you sign up to be notified when he uploads new data sets.

In addition, he doesn't claim to have done the underlying work. The ideas and implementation are his, but he's quick to credit decades of work by those who preserved, photographed and indexed these records.

Luke is working on adding new data for both records and finding aids. I gave some ideas of data sets that might lend itself to his site, and he welcomes additional feedback through the site itself. He also shared what's coming soon:

Ready, awaiting release

• Pogrom relief and Jewish resettlement files, Ukraine 1917–1938. About 502,000 people named on 56,900 pages from the Kyiv oblast archive and Ukraine’s central state archive (TsDAVO). They come from pogrom relief committees, the Jewish land-settlement bodies (OZET and Ukrkomzet), government ministries, and the protocols of the “troikas”.
• Jewish name lists of the Russian Empire and Ukraine, 1851–1943. 22,766 people from thirteen books of lists from Ukrainian archives, shared on Wikisource by Krakovsky. They include:
  • Jewish army recruits of 1853–1860
  • jewellers’ registers of 1893 and 1899
  • emigrants to America through Galveston, 1910–1914
  • victims of the 1918–1921 pogroms
  • refugees at the Romanian border in 1921
• Conversions in the Russian Empire, 1796–1915. About 33,000 people from the files of the Kyiv, Podolia and Zhytomyr consistories and the governors of Kyiv, Volhynia and Chernihiv, including 4,539 converts from Judaism.
• Galician Conversions, the new name for Lwów Conversions. It gains 13,471 people from converts’ books and civil registers in Lviv (TsDIAL), Warsaw (AGAD), Kraków and Przemyśl, for 54,602 people in all.
• More Galician registers: Trzebinia 1880–1911 and late birth entries from Żołynia, about 1,500 records.
• Swedish Jewish congregations, 1886–1911: member books, migration lists and births from Kalmar, Oskarshamn, Växjö and Sundsvall.
• Sibiu, Mediaș and Alțâna (Transylvania) Jewish registers, 1854–1944: about 5,000 births, marriages, deaths and requests.

In progress

• Lithuania, interwar Jewish community registers, 1911–1940. Marriages, deaths, divorces and births from 63 communities, including Vilnius and Kaunas, from the Lithuanian State Historical Archives. That’s about 50,000 images in Lithuanian, Yiddish, Hebrew, Russian, German and Polish. Reading has started, Vilnius and Kaunas first.
• Russian Empire case files, 1880–1919. About 52,000 pages from 847 files of the Kyiv governor’s office and other archives: Talmud Torah reports, prayer-house and rabbi elections, expulsions, residence petitions and pogrom files. The pages that name people are being read.
• Volhynia and Rivne Jewish community lists, 1930s, and Kyiv synagogue registers from 1893. Typed tax and census lists of 1937–1939 with addresses and occupations, plus later alphabetical registers.
• Older Russian Empire Jewish registers, before 1868, are being tested.
• Still waiting for better reading: Congress Poland registers of 1868–1917 written in Russian, Galician registers from before 1870, and pages in Hebrew letters.

Luke is constantly adding additional features as well. Today he added additional functionality, a page which lists everything that has been indexed.

I admit that part of me mourns the loss of the 20-hour hunt that ends in a find. But last night, a marriage record that took seconds to locate on archival.name reached a that couple's 99-year-old Holocaust survivor son who thought nothing of his parents' remained. That is what this is about.  And I can't wait to see what else is to come as we integrate AI into genealogical research.

You can like my page on Facebook:

or follow @larasgenealogy on Twitter/X.

Receive new posts by email: