Now Indexing — Again

 

There's been a lot of talk lately about new versions of Artificial Intelligence (AI) in its latest incarnations, such as ChatGPT. But in the midst of all the breaking news on AI developments, I forgot one quietly advancing employment of such AI systems that we genealogists can appreciate. 

Remember how FamilySearch needed so many volunteers to "index" the many records they were bringing online on their website? They had so many more records digitized—but in browse-only format—than they could make available in a searchable form, since it used to take human eyes and hands to type in the words captured in the pictures scanned from billions of records from around the world.

And then, remember how, with the release of the 1950 census, organizations like FamilySearch and Ancestry.com put a version of AI computers to work indexing those records, accelerating the time it took to prepare the scans for release to the public in a searchable form?

With the advent of computers trained to recognize handwriting, it seemed there would soon be no more use for volunteers to complete the labor-intensive process of transcribing information from the pictures taken of original documents—what FamilySearch calls indexing. And while we were glad to help, I imagine there was a collective sigh of relief in realizing a brighter day was peeking up over the family history horizon with our first glimpse of what AI could do with the 1950 census.

Truth be told, after I participated in the April all-hands-on-deck indexing fest working to double-check what the AI machines offered as a first draft, I figured my volunteer hours were done for good. From this point on, AI would take the stage—and shine.

While that is not entirely correct—there are still projects pleading for volunteers to help—a recent news article reminded me that AI is indeed gaining ground in this project to convert old microfilmed records to searchable, digitized formats. According to Rebecca Olds' article this past month in Deseret News, last year FamilySearch finished digitizing more than sixty years worth of microfilmed records. In her interview with senior production manager John Alexander of FamilySearch, he claimed that within a couple hours of work, a trained computer could index more documents than any one volunteer could index "in a whole lifetime."

The article, "How FamilySearch is Using the Future to Discover the Past with AI," gives us a sneak peek at the processes going on behind the scenes as FamilySearch brings decades of governmental and private records to the forefront for those of us seeking information on our elusive ancestors. Everything from details on how the computers are trained to do the work of transforming those microfilms to reports on the sheer numbers of records yet to be made available are, in my opinion, mind-boggling.

Though the science behind this progress is impressive, there is admittedly much more to be done. As the article reminded me, volunteers are indeed still an integral part of this process. As FamilySearch's production manager John Alexander mentioned, "computer automation isn't going to replace our volunteer indexers—if anything, we need more of them."

Guess it's high time I got back to work.