Thursday, March 9, 2017

Is There Really an Ancestry.com "Big Tree?"

Many users of the We're Related mobile app are puzzled by having generations added to their ancestral lines and having common ancestors that are not in their own family tree.  Why and how does that happen?  I'm finding that about 30% of my "cousin relationships" are not realistic because of this - and as a result I don't trust any of the relationships provided.  They say it is for fun only, but fun things need to be accurate.

Some readers might say "well, you don't have those generations of ancestors because you haven't done a good enough job of researching your ancestry."  That may well be.  However, I am pretty sure I have almost all of my ancestral lines correct back to either an immigrant ancestor or an ancestor for whom I have been unable to find parents.  The app has yet to find a correct set of parents for my "end of my known line" ancestors that I can find a record for.

1)  When the user signs into the We're Related app, we are prompted to add ancestors or connect to an Ancestry Member Tree.  The app shows only three generations of ancestors for the user.  Why doesn't it show more than three when my tree has more generations?  I think the answer may be this:

*  Ancestry.com may use the eight great-grandparents in a person's tree to hook into a "Big Tree" that Ancestry.com has developed in order to enable finding the We're Related cousin relationships.

*  The "Big Tree" may be a consensus approach, based on entries in many (or all?) of the Ancestry Member Trees in the system.  

2)  We see a consensus approach when we click on the DNA Instant Discoveries - when you click on the person you get a statement that the profile is "Compiled from XX Family Trees."  There is a question mark next to that that shows this graphic:


On those consensus profiles there are often several spouses and dozens of children, as one might expect when someone combines profiles from a number of individual trees.  One of my Instant Discoveries is in 77 trees.

On the graphic above, there is a link for "The science behind the compiled tree."  :


The graphic above says the factors that contribute to a positive "quality score" are:

"• Completeness of information. Complete vital information (given name, surname, gender, birth date and place, death date and place, marriage date and place).
Number of attached records. Ancestry members can attach records to a person as sources for various genealogical events and facts (for example, a birth certificate can be attached as evidence of a birth date or names of parents). The more records that are attached, the higher the quality score.
Quality of the records. Certain types of historical records offer better genealogical evidence than others. For example, a birth or death certificate offers more information and is more likely to be correct than other records such as city directories. So the quality score increases when a fact has strong genealogical evidence to support it.
Number of attached media items. Ancestry members can attach their own media items such as photos and stories to a person. Because media items may provide evidence for the vital information, events, or relationships of an individual, they also increase the score."

That's interesting.

3)  Then I read the Ancestry Insider's blog post "Passionate Genealogist is Core - Tim Sullivan at #RootsTech" and he paraphrases Tim Sullivan, the CEO of Ancestry.com, saying:

"...Tim went on to explain more about their Big Tree. Big Tree is an internal term they use for an effort they’ve been engaged in for many years to stitch together the millions of member trees on Ancestry.com. They are applying machine learning technologies and authority systems and are getting more accurate every day.

“'That has always been a little bit of a holy grail, to find that one tree,' Tim said. They are taking a different approach than FamilySearch, but then again, their purposes are different. The Big Tree is not intended to be a product, but something that allows them to develop 'some pretty cool capabilities.' One application is their new We’re Related app."

So there is the explanation - there IS an Ancestry.com "Big Tree," and it likely represents the "Compiled View" mentioned in the DNA Instant Discoveries graphic.  

5)  I haven't found a way to access this "Compiled view" other than by using a Google search.  I did a search for ["william knapp" 1775 1856 ancestry.com] and quickly received this graphic (three screens):


And there are the erroneous parents for my 4th great-grandfather, William Knapp (1775-1856).  The purported parents married in 1783, and the mother was age 4 for her purported son's birth.  How did the "Quality Score" miss that?

6)  In my Friday series of We're Related posts, I have found that about 30% of my lines back to the purported common ancestor are wrong - just flat out wrong - because of errors made by the "quality Score" in the "Compiled View" Tree.   

7)  If Ancestry is going to have a "Compiled View" tree and base "cool capabilities" on it, then:

*  they need to be willing to improve it and remove the obvious errors created by unrealistic relationships, unrealistic number of children, same name-same place problems, same name simultaneously in different places, etc.  
*  there needs to be a "Discussion" feature and/or a review board of experts who can evaluate disputed relationships and information.  

8)  Frankly, just because there is a consensus from N different sources about a fact or relationship, that doesn't mean the assertion is correct.  We, as genealogists and family historians, are really in the business and avocation of developing accurate and complete person profiles, relationships and life events.

It would be terrible to disappoint my grandchildren or your grandchildren when they find out that they relied on the We're Related app, and then they found out they weren't really related to Genghis Khan, Barack Obama, William the Conqueror, or even Beyonce.

I know that Ancestry.com had a "One World Tree" back several years ago, but it was taken off of the database list.  Is this "Compiled View" just an expanded "One World Tree?"

Perhaps Ancestry.com will expound more on how they put this tree together, how they review it and correct it, and how users can access it other than by Googling for a person (that does work well!).

I found this article by Googling a bit:  Where Do We Come From? Ancestry.com’s 19B-Record ‘Big Tree’ Database May Have the Answer 

UPDATED to fix typos and clarify sentences.

==========================================

The URL for this post is:  http://www.geneamusings.com/2017/03/is-there-really-ancestrycom-big-tree.html

Copyright (c) 2017, Randall J. Seaver

Please comment on this post on the website by clicking the URL above and then the "Comments" link at the bottom of each post.  Share it on Twitter, Facebook, Google+ or Pinterest using the icons below.  Or contact me by email at randy.seaver@gmail.com.


4 comments:

Michigan Girl said...

Thanks Randy. Excelllent observations and information. I too wish we could click on the info presented by the We're Related app and find the tree those supposed ancestors belong to.

Marian said...

It seems to me that Ancestry could improve the quality of its "quality score" greatly by just doing what we do: discount the sources that turn out to be other member trees or "Gedcom." They still would have the embarrassing children who were born before their parents, but they'd eliminate a lot of non-sources in one step.

Stephen Mumford said...

Hi Randy,

I am a long time follower of your blog and visit on a regular basis. I had previously used your same approach of a Google search to try and access their "Big Tree"

Today I noticed that even in your screenshot there was a reference to "Historical Person Search" in the breadcrumbs at the top of the page.

This takes you to:
https://www.ancestry.com/genealogy/records/

It is not as flexible as a Google search but it does give you a partial interface to this data.

Randy Seaver said...

Stephen, I had missed that link. I see it now.

Strange thing: I searched that link for my William Knapp (1775-1856) and there was no link. I did a search on Google and the link for him in the "Historical Person Search" came up.