Holocaust researchers use AI to search for unnamed victims

FILE PHOTO: The “Arbeit macht frei” (Work sets you free) gate is pictured on the site of the former Nazi German concentration and extermination camp Auschwitz, two days before the 76th virtual anniversary of the liberation of the camp in Oswiecim, Poland, January 25, 2021. Picture taken January 25, 2021. REUTERS/Kacper Pempel/File Photo

By Steven Scheer

JERUSALEM (Reuters) – Researchers in Israel are turning to artificial intelligence (AI) to comb through piles of records to try to identify hundreds of thousands of Jewish people killed in the Holocaust whose names are missing from official memorials.

FILE PHOTO: Visitors tour an exhibition, ahead of Israel’s national Holocaust memorial day at Yad Vashem, the World Holocaust Remembrance Center, in Jerusalem May 5, 2024. REUTERS/Ronen Zvulun/File Photo

More than six million Jews were murdered by the Nazis during the Second World War, a genocide commemorated across the world on Monday on Yom HaShoah, or Holocaust Remembrance Day.

In the build-up to those commemorations, staff at the Yad Vashem World Holocaust Remembrance Centre in Jerusalem said they were working to step up searches for details of known and unknown victims after developing their own AI-powered software.

FILE PHOTO: A visitor looks at pictures of Holocaust victims at Yad Vashem’s Holocaust History Museum in Jerusalem April 18, 2012. REUTERS/Nir Elias/File Photo

Over the years, volunteers have tracked down information on 4.9 million individuals by reading through statements and documents, checking film footage, cemeteries and other records.

“It’s very hard for a human being to do it – just to go over everything and not miss any details,” Esther Fuxbrumer, head of software development at the centre, told Reuters.

There are huge gaps in their existing 9 million records. The Nazis “just took people, shot them, and covered them in a pit. And there was no one left to tell about them,” Fuxbrumer said.

And then there is the mammoth task of linking individuals to dates and family members and other details, watching for duplicates and comparing accounts.

The AI system, developed over the past two years to sift through records in English, Hebrew, German, Russian and other languages, is currently undergoing tests.

“That technology works very fast, it takes a few hours to go over hundreds of testimonies and it’s very exact, the results,” Fuxbrumer said.

“We saw that from each testimony we could get between six or seven names with full details that we could put automatically into our database, and around 10% of the names that we found we had already in our database, but 90% are new names that we didn’t know about.”

In one case, they found information about Yehudit and Ruth Rosenbaum, two twin four-and-a-half-year-old sisters from Romania who were taken to Auschwitz. Yehudit survived. Ruth was murdered.

“And we were able to bring more information about Ruth from someone that’s not her family at all, someone that met her in the camp,” said Fuxbrumer.

In the trials, staff are running tests on 400 of their 30,000 testimonies, including many three-hour-long recorded videos of survivors.

Fuxbrumer said 1,500 new names had been added, and many more were expected in the coming weeks, when the system is used on all 30,000 testimonies. The next phase of the trial will cover diaries.

“We believe that way we’ll be able to bring a lot of stories about a lot of victims that were killed, little kids that no one else knew, to tell us the story about what happened to them.”

(Reporting by Steven Scheer; Editing by Andrew Heavens)