Monday, 2 April 2007

Family Search Indexing

There's been a lot of buzz about Family Search Indexing, an initiative from the LDS Church. They're looking for volunteers to extract family history information from digital images of historical documents. This will create indexes that "assist everyone in finding their ancestors." The web site boasts "in 30 minutes you can help people find their ancestors!" I thought I'd try it.

Before you can index you have to register and download the special application. It runs independent of your regular browser. Registration is online. You don't have to give personal information -- but you do have to provide an email address. When I registered it took a day for them to get back to me with the password, likely too long to be able to use one of those temporary email addresses.

The application download is 28.6MB, not something you want to try on dial-up. Installation went smoothly and, after changing the initial password, I was ready to download the first batch of images. You have a choice of projects to work on. See the list here.

First I tried a page of the 1900 US census from New York. The screen is divided into two, the original image on top, the transcription you are making below. They attempt to highlight the element you should be transcribing, and usually come close.

It's good to read the instructions first so you know what to do when anomalies arise, such as missing data or entries that don't match the default format. Failing that information is repeated for each element as you come to it in a small box to the lower right.

If you're reasonably computer literate it's reading the original document image that you'll find to be the greatest challenge. Even with a good capability to magnify it can be tough. No doubt that improves with practice, I didn't feel I'd anywhere near mastered the skill.

For the second batch I switched to transcribing Ontario death registrations and felt more comfortable as I have a better grasp on the geography. The data wasn't as compact as on the census and you got a second chance at the name as there was repetition.

It turns out the images I dealt with had already been transcribed by Ancestry, although not in as much detail. It was certainly tempting to check what I read against the Ancestry transcription. I tried it a couple of times and didn't feel I did any better a job. It's also tempting where you find something difficult to read to search for a complementary record, like a census, to resolve the problems. In most cases where I thought I found one it was hard to justify using the spelling from that other record. You might question whether that one is correct anyway.

I couldn't complete a batch transcription in 30 minutes, which seemed to be the promise. Maybe I'm slow and will get there with practice. You can always work for 30 minutes, save the result and come back later to continue the task.

At the end of the batch the program allows you to run through and review elements it finds unusual. I had quite a few in each batch, some for missing elements, some for names that were non-English. Apparently the same batch is served to two different people, then a third looks at items where those two differ.

Given that the data I was indexing was already on Ancestry, for which I have a subscription, I found my motivation flagging a bit. You can go in and see a list of the items you've completed but it would be nice to receive some additional feedback. Part of the problem was that there were no records on offer that I had a personal motivation to index. Presumably that will change as the project gets up to full steam.

My recommendation? Why not give it a try. We all benefit when more data becomes available. Even if it's already indexed the competition helps keep the commercial guys honest.

No comments: