Friday, February 6, 2009

Wrong images in FamilySearch Record Search

I received this inquiry back at the end of October 2008:

On June 13, 2008, I wrote to Family Search:

In the Texas Death Records 1890-1976 database, the index data for film #2116470 seems to have been mismatched to the images for film #2134914.

Here are two examples to illustrate the problem:

Search for "Earl Jeff Smith," who died in 1935 in Harris County, Texas. Data indicates film #2116470, image #2211. Display image and you get the death certificate for "Charles Gilbert Baker" who died in 1929 in Colorado County, Texas. Search under Baker's name and data says film #2134914, image #2211.

Search for "Harvey S. Smith," who died in 1935 in Harris County, Texas. Data indicates film #2116470, image #2300. Display image and you get the death certificate for “James Thomas Heflin” who died in 1929 in Dallas County, Texas. Search under Heflin’s name and data says film #2134914, image #2300.

I’m trying to get to the image for “Earl Jeff Smith,” and would be grateful for your help in rectifying this small glitch in an otherwise terrific online resource.

Ellen Dunlap

My response in early November was:


Thanks for the examples. I'll see if I can find out what the problem is and the outlook for a solution.

I can tell you that watching the team work on problems, it feels like I'm watching a pit crew trying to change the tires on a race car--without pulling the car out of the race.

The team is more heads down on fixing the process problems that created these data problems than they are on fixing the data problems themselves. That's one reason this is called a pilot. It doesn't make any sense to try and send the collection through the pipe again if it is just going to spew out bad data again. And it makes even less sense to pull workers off fixing the pipeline to fix a single collection by hand.

I'll let you know when I hear something.

-- The Insider

Ellen's been patient. But it's been three months since she wrote me and seven months since she wrote FamilySearch, so she recently asked for an update:

Did you ever hear back on this matter we discussed in November?  The data/image mismatch problem still exists.  If two films -- 2116470 and 2134914 (not even close enough to be a typo) -- are completely switched, that sounds like a process problem that might be replicated with other films as well.

I have written to FS also, but had no reply.



Dear Ellen

I'm sorry you never heard back from me. This is one of the cases where "no news is bad news." I haven't forgotten your request, however. Just recently I heard a list of collections read for a project re-processing some missing images from microfilm back at the Granite Mountain Record Vault. I was listening for, but didn't hear "Texas Deaths, 1890-1976" mentioned.

Films Swapped?

Under the assumption that the two films (2,116,470 and 2,134,914) were completely swapped, I tried to find the name to search for that would give the corresponding image from film 2,134,914.

I searched for Earl Jeff Smith (of roll 2,116,470) as you did and found the certificate for Charles Gilbert Baker (of roll 2,134,914). I searched for Baker hoping to get the certificate for Smith. Unfortunately, I got the certificate for Baker. The two sets of images have not been swapped. Rather, the images of one set have been swapped.

Duh! Re-reading your message, Ellen, of course that is what you meant. My computer science brain was interpreting your message using the semantics of a computer scientist rather than a normal person.

Process Problem?

'Swivel chair' processing was required where gaps existed in the pipeline
'Swivel chair' processing was required
where gaps existed in the pipeline
Credits: Pipeline, Chair
Modifications © 2007, The Ancestry Insider

As a result, all images for film 2,116,470 are wrong. As you mention, that certainly sounds like a process problem. I asked about this and was told that yours is one of several wrong-image problems in the "Texas Deaths" collection that were introduced by swivel-chair handling rather than digital pipeline processes. While still working at, I did a four-part article on the FamilySearch Digital Pipeline, based on information presented at a BYU genealogy conference.

In part 2 I explained that swivel-chair processing—manual processing—was required where gaps in automated processing existed. In the case of your "Texas Deaths" image problem, I guess this means that creating the sets of images and copying them into the correct places, was a manual process. Now that step has been automated and the automated process doesn't suffer from the human fallibility that created your problem.

So when will you be able to get Earl Jeff Smith's death certificate?

That, my dear Ellen, in my opinion IS a process problem. Speaking as an industry expert without regard to what internal knowledge I may have about FamilySearch, these are some processes that an archival-quality vital records repository ought to have:

Stay tuned for "Standards of an archive-quality digital record repository."

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.