Statistical and summary data of those who died in the 1918 influenza pandemic has been available in various forms for quite awhile. But not their names.
The Brigham Young University Record Linking Lab is developing one of the first data sets of each individual who died in the 1918 pandemic by extracting the cause of death from death certificates.
“There have been people that have been doing it by hand, kind of going through death certificates, looking for people that died of influenza and pneumonia,” said Joseph Price, a BYU economics professor who oversees the lab.
However, “this is the first time that it’s been done in a fully automated way and on such a large scale,” he said.
The lab links the records of 1918 pandemic deaths to FamilySearch’s Family Tree, where a profile for each individual has been created.
“This does two things: One, it allows us to link these people to their family members on the Family Tree so we can study what was the impact on the pandemic for the survivors within the family,” Price told the Church News.
Second, “it provides a natural place for people to gather and share the memories of these people so that we can draw on them for strength during this pandemic and future pandemics.”
The data can be particularly helpful to those looking to study long-term effects of the 1918 pandemic — and learn from and apply that information to the current pandemic. By comparing the 1918 pandemic’s trajectory with the trajectory of COVID-19, researchers may better understand the effectiveness of public health interventions in reducing deaths.
Data from 10 states — including Utah, Idaho, Texas, Ohio and Massachusetts — is now available in an interactive map at pandemic.familytech.byu.edu. Another eight states are in process, with the goal to eventually include all 50 states.
How the automated process works
The death certificates were previously indexed by FamilySearch. “What we focus on is the teaching machine to look for where the cause of death should be and then reading that handwriting,” Price said.
The BYU Record Linking Lab’s process involves deskewing (making the record image straight), segmenting (teaching the machine the location of the desired fields) and handwriting recognition to read the information.
Prior to the COVID-19 pandemic, Price and others had been working on a project to extract the cause of death specifically for Ohio death certificates. “When the pandemic hit, my lab realized we could repurpose those tools to focus specifically on the 1918 pandemic and then expand on to other states,” he said.
To create the dataset, students in the lab identified and retrieved hundreds of thousands of relevant images from FamilySearch.
Eric Burdett, student project leader, wrote computer code to interface with FamilySearch’s system. “That’s been quite a process because their collections are just massive,” he said in a BYU news release. “We have access to millions and millions of records from FamilySearch, resources a lot of researchers haven’t had before.”
The data available in the interactive map currently includes deaths only in 1918. “But the tools we’re using could be expanded out to people in 1919 and 1920, who were also victims of the pandemic,” Price added.
Why it matters
To Price, the 1918 death certificates are one example of historical records that have modern relevance. The “bigger picture” is looking at the value of auto-indexing tools and making more records searchable.
“The same technology that unlocks the cause of death is going to also read your grandpa’s journal,” Price said. “It’s going to extract additional fields from census records. It’s going to pull additional little details from other records.”
Ultimately, “All the time and energy we put into solving this particular problem is now helping us solve other problems that are going to unlock records and allow people to find their family and do their temple work. And that’s pretty exciting.”
Read more: How FamilySearch is using computer technology to speed up the process of publishing record images
Price said he recently used tools to auto-index his father’s mission journal. “It kind of opened my eyes to what’s going to be possible as all of us have journals, records, documents, sitting in our homes, sitting in our libraries, in our archives. Those records all have ‘needles’ that people are looking for — like we’re always looking for needles in haystacks.
“And so I’m very excited when all those needles are fully searchable. And you’ll be able to gather all the results together in one place and not have to go and search through every haystack.”
To solve the problem of making every record in the world searchable,“I think it’s going to be a combined effort across a lot of groups,” Price said. “And handwriting recognition can be a real key part of that because so many records are handwritten.”