Friday, March 25, 2011

Exploring Wikitree - Post 9: Another Search for Matches

Previous posts in this series are in the summary post The Exploring WikiTree Compendium.

One of the possible uses of any online family tree system is that it be searchable for potential common   ancestors submitted by the registered users of the website.  A successful search may result in communication and collaboration between researchers to the (hopefully) mutual benefit of all involved. 

In Post 8, I mentioned that I had not found any potential matches in the WikiTree system to my profiles, but I searched manually, and only casually, using the Search fields at the top of the WikiTree page.

There is another way to search for Matches:  On the user's "NavPage," there is a link to search for potential matches.  Here's my NavPage:



The link to the "Search for Matches to see if your tree overlaps with others" feature is just above the small To-Do list box which is just above the images I've uploaded at the bottom of the screen above.  I clicked on the link and saw the "Find Possible Matches" screen:



On this screen, I entered the surname "white" in the field.  The Note says: "This search tool is brand new. It will return many 'false positives,' i.e. suggest matches that are incorrect. We will be releasing improvements to the matching system soon."

Fair enough, at least it searches the WikiTree system.  Here are the results for White (two screens shown, but six pages for White):




At the bottom of each page is a button to "Look for more matches."  This Search feature seems to take every person with the White surname in my tree and finds persons with the same given name and surname in all of the trees on WikiTree.  The matches include my own WikiTree entries that do not match exactly.  The information provided for each potential match includes:

*  Name
*  Birth date
*  Death date
*  Profile manager (with a link)
*  "Merge" link

There are colored buttons next to the names.  I've seen open, green and red so far.  I don't know exactly what they mean.  Logic indicates that red would mean "definitely not a match" and green would mean "a potential match."  There must be some algorithm in play here.

I did find one possible match - for Lydia White born 10 May 1686.   I wonder why this match, which has the name and birth date correct, isn't highlighted with some sort of "exact match" button?  Perhaps it's because the birthplaces and the death dates didn't match.  Here is the profile for my Lydia White:


Here is the profile for the found Lydia White:


I requested to be on the Trusted List for this Lydia White.  If I am accepted, then I can merge my profile with the matching profile.  Collaboration!  It works, if people play in that sandbox.

Since I did not have ancestors of my Lydia White in the GEDCOM I uploaded (but I do have them in my database), I may upload a small GEDCOM to cover this line in order to help this researcher.

One thing I noticed when I was going through the Profiles for the "end-of-line" ancestors in my WikiTree, like Lydia White, is that the siblings are listed but not the parents.  Which means if I upload a new GEDCOM for, say, the ancestors of Lydia White, I will have to merge the two Lydia White Profiles that I've uploaded, and probably will have to merge the siblings also.  It may be easier to upload a GEDCOM with the parents, without their children, and then connect my Child to the Parent.  We'll see how this works out!

3 comments:

Chris Whitten said...

Hi Randy,

Our search system is more primitive than you imagine! Good search algorithms are remarkably complex. We're releasing some major improvements to our matching system in the next couple days, but it will still return a lot of matches that any human could easily eliminate.

The colored icons don't indicate anything except the Privacy Level. I'll look for a way to clarify that.

Chris

Shane said...

I think this comment is more for Chris than you, Randy, but maybe you'll have some insight, too.

One difficulty is that if you aren't on a trusted list for an individual who matches yours, you can't merge even if you want to. It would be nice if I could mark all those results as a possible match, and others as not a match, so next time I search for matches (assuming more people have added to wikitree), I don't have to sort through all of the cruft again. And so that I can look through the possible matches more closely and merge once I'm added to a trusted list.

Chris Whitten said...

Hi Shane!

I like your idea of having some sort of "this is not a match" indicator. This could be public information displayed somewhere. It's valuable information for anyone looking for Joe Schmoe that a WikiTreer has compared this other Joe Schmoe and believes they are not a match.

This will take some work, but it's going on the to-do list. E-mail me or let's start a Facebook discussion for it if you have more ideas.

Here's something related that's coming soon. We have a "merge proposal" system that will be released within a few weeks. This will enable you to skip the step of joining the other Trusted List. You can propose a merge, and then the other Profile Manager can approve it. This tool will then enable other collaboration-enhancing tools to flow from it.

Chris