Now Indexing : Something a Machine Started

 

It's been my habit every month to do my part to "give back" by volunteering to index digitized records at FamilySearch.org. Believe me, I'm not alone; there are "hundreds of thousands" of folks around the world helping with this ongoing project. The problem is, there is an avalanche of nearly forty million records being digitized each year, through over three hundred cameras operated by FamilySearch in fifty countries around the world. It's hard to keep up, even with hundreds of thousands of indexing volunteers.

That pace, incidentally, is accelerating: just last month, FamilySearch reported having indexed thirty four million records from around the world. That, in part, is the result of a new approach to converting pictures of records into readable—and thus searchable—documents through indexing. Instead of approaching the task entirely by hand, FamilySearch is now employing not only the newer Optical Character Recognition systems, but combining that with machine learning and artificial intelligence to extend that character recognition capability to handwritten documents, as well. A demonstration of how that process works was provided at last month's RootsTech Connect, where you can actually upload a sample from your own ancestor's letters to see a demonstration.

For the greater part of history, the types of documents we have relied on to verify our family history were handwritten. Developing a system—even if it still needs the intervention of humans double-checking computer work—will yield far faster results than we've ever seen before this point. This is indeed encouraging news.

But what is it like today, when I sign in to my FamilySearch account to index records? I've been watching the choices available as indexing projects, and the list seems to be much shorter than I remember it in previous years. 

For this month, of the few projects in English in my home country, I found one on passenger lists arriving in the port of New York City, and decided to work there, remembering how helpful those lists were in my recent project to trace my godmother's father, a career seaman. 

Other than reading the step-by-step instructions and spotting the warning that the scanned records would be faint or hard to read—oh, my eyes!—there was little to do but move through the brief steps on a short batch of records.

Once completed, I could see why this indexing batch might not have been read by computers: the scans were faint, dark, and poorly focused. While computers may be speeding up the overall progress of transforming digitized documents into searchable records, the few remaining images they can't handle will take a disproportionate amount of an indexer's time.

Overall, while the news of this computerized progress is great, I may have to re-evaluate my usefulness to the program, if for nothing else than my eyes' inability to decipher the rejects those brainy computers can't handle.