Thursday, February 17, 2011

RootsTech: Future Directions in Search

These are my notes from the session “Future Directions in Search,” by Ian Tester, a product manager at findmypast.co.uk, a brightsolid website.

Find My Past logo As an aside, Tester mentioned that Find My Past will shortly be adding records from Ireland.

Ian Tester of Find My Past said the input to a search is a little bit of information about an ancestor, and the output is an ancestor. (Wrong, Ian. The output of a genealogy search is not a person, but a record. You’re committing the mistake that I’ve so often criticized in this space. Genealogy is deceptively complex. Sources, information, and evidence are distinctly different than conclusions, ancestors, and trees. The family history community badly needs tree management applications that include evidence management. But I digress…)

Lots of search fields with no meaning to foreigners The 1901censusonline search form has lots of search fields that have no meaning to foreigners and some of little meaning to Englishmen, said Tester. This could be simplified to who, where, when, what, and with whom.

“It’s not just about having three fields instead of 20,” he said. Users need better, more intuitive ways of specifying who, where, and when. Why did the GRO stop recording middle names, then start again? The search experience should adjust to user input accordingly.

Find My Past is doing experiments on identifying connections between records, such as matching the bride and groom in marriage records. It will give you the likelihood that a couple is a match.

Another possibility is matching of family groups between censuses using algorithmic rules. Tester said they worked with genealogists talking to developers to teach computers to think like family historians. “Although we were doing it as an experiment, we were amazed at the efficiency that the machines could do.”

“You don’t want to return results so exactly that you eliminate the serendipitous result.” (Ian, I think there is room for both. That is why the user should be able to choose between exact and ranked results. Oops; I’ve digressed again.)

If you’re going to return matches from user submitted trees, you better be explicit about doing this and do it sensitively. (Ian, how right you are!)

A lot of what the industry has been doing Tester calls “digitize and dump.” “When you put data online you don’t have to leave it as static data,” said Tester. “You can nurture it and grow it. Use the power of your users—human intelligence—to improve the data.” (Ian, you are spot on.)

When Tester asked for questions, the first person didn’t ask a question, but made a list of suggestions. Ian loved it, saying they were super ideas.

That’s what RootsTech was all about.

5 comments:

  1. Hi Insider,

    First of all thanks for coming by, and covering my talk, it was really great to be at Rootstech. Apologies for any oversimplifications in my talk, especially around search "outcomes" - we are very focussed on delivering evidence-based outcomes to users and this concept is at the heart of findmypast's DNA - as we originally came from a Forensic genealogy background, primary sources and the accuracy of them are entirely at the heart of what we do. I'd be delighted to talk further on any aspect of any of the issues i raised - it'd be a pleasure to talk face-to-face the next time our paths meet. Go Rootstech!

    ReplyDelete
  2. Thanks for your coverage of RootsTech. Your blog has been the most informative of any.

    ReplyDelete
  3. Ian Tester of Find My Past said the input to a search is a little bit of information about an ancestor, and the output is an ancestor. (Wrong, Ian. The output of a genealogy search is not a person, but a record. You’re committing the mistake that I’ve so often criticized in this space. Genealogy is deceptively complex. Sources, information, and evidence are distinctly different than conclusions, ancestors, and trees.

    This is why I contacted you about the bulk merging on family search. Something needs to be done quickly.
    Suzanne

    ReplyDelete
  4. Good notes. One thing you didn't mention that I got excited about was the mentions of expanding the geography of a search. We all know that someone recorded as living in county X will often have records (of them and their families) in surrounding counties, even across state lines. (or replace these terms with parish, shire, country, etc) With geocoding, such searching would be very practical, and tremendously useful. After suggesting the idea (not very effectively) for a decade, I'm still not seeing this searching available from the vendors. Ian went one step further, suggesting that geographic, economic, and cultural boundaries could be considered in such searching. I WANT it - for censuses, familysearch.org, even tree databases. Best part of a very good session!

    ReplyDelete
  5. Repeat after AI: "Sources, information, and evidence are distinctly different than conclusions, ancestors, and trees."

    This should be emblazoned on that nice iceberg graphic from California.
    Thanks for your continuing emphasis on this point despite enduring seeming obliviousness to it in many quarters.

    ReplyDelete