Wednesday, September 19, 2012

First Look at Record Matches on MyHeritage

I took a spin through my Record Matches on MyHeritage today (see announcement and demonstration at Introducing Record Matching. 

In order to perform these tasks, I have to have a MyHeritage family tree (I do!) and a subscription to WorldVitaslRecords.com (I do!).  The way the Record Matches work is that MyHeritage looks at persons in my MyHeritage family tree, tries to find them in the free or subscription record collections that they have access to, and if they find a match they add it to the Record Matches list.


1)  Here is the Home page my MyHeritage website:



2)  I clicked on the "View your Record Matches" link on the screen above, and went to the Record Matches page on the Family Tree tab:


The page above lists all of the record collections in which the Record Matching technology found matches.  There are 3,765 Record Matches for people in my tree, and they are in 30 record collections.  The collections that are listed for my Record Collections are:

*  1860 U.S. census (7 matches)
*  1940 U.S. Census (8 matches)
*  U.S. Air Force Register Extracts (5 Matches)
*  California Births, 1905-1995 (227 Matches)
*  California Deaths, 1940-1997 (234 Matches)

*  Data Relating to the Settlement and Settling of New York and New Jersey (3 Matches)
*  Early Germans of New Jersey (4 matches)
*  England & Wales Deaths, GRO Indexes, 1969-2007 (46 matches)
*  Everton Pedigree and Family Group Sheets (9 matches)
*  Find-A-Grave (1,322 matches)

*  Historical and Genealogical Miscellany, Volume 1 (3 matches)
*  Illinois Deaths, 1916-1950 (4 matches)
*  Illinois Deaths before 1916 (1 match)
*  Illinois Marriages, 1763-1900 (50 matches)
*  Kentucky Births, 1911-1999 (1 match)

*  Kentucky Deaths, 1900-1999 (2 deaths)
*  Leicestershire Parish Birth and Christening Records (2 matches)
*  Leicestershire Parish Marriage Records (3 matches)
*  Maine Deaths, 1960-1996 (3 matches)
*  Maine Marriages, 1892-1966 (24 matches)

*  Maximilian Family Tree (157 matches)
*  Membership of the Church of Jesus Christ of Latter-Day Saints, 1830-1848 (4 matches)
*  Newspaper Archive (818 matches)
*  Social Security Death Index (738 matches)
*  Texas Births, 1926-1995 (53 matches)

*  Texas Marriages and Divorces (29 matches)
*  The Early Germans of New Jersey (2 matches)
*  The Early Germans of New Jersey, Their History, Churches and Genealogy (2 matches)
*  U.S. World War II Army Enlistment (1 match)
*  UK Marriage List, 1655-1992 (1 match)

3)  The default filtering option seems to be 1 star (whatever that means).  Up under the number of matches, there is a link for "Filtering Options" (273 matches not listed).  The user can choose how many Stars to select - 5 stars gave me 48 matches in 5 collections, and 0 Stars gave me 4,038 matches in 32 collections.  My guess is that most users won't modify the default filtering.  Here is the Filtering window:


On that window, the user can choose the group of matches to display (e.g., pending, confirmed, rejected) and the Record types to display (e.g., structured or text).

4)  I wanted to look at how the matches are presented.  Here is the 1940 U.S. Census page for my 8 matches):



In the screen above, the information in my tree is on the left side of the screen (birth death, parents, spouses, children) and the record data is on the right side of the page (birth, residence, parents, spouse, children in the record).  I can click on the blue "Review Match" button for each match to see the record, or I can click on the check mark to confirm the match, or the X mark to reject this match.

I clicked on the blue "Review Match" button for one of the 1940 Census matches (3 screens below):




In this record collection, the information in my tree is at the top left of the screen, and the information about the person in the record is on the right side of the screen.  The census source information is provided (NARA roll, image frame, state, county, township, enumeration district, page, family number andl ine number), but no source citation is created).  The record image is provided, and can be manipulated within the frame, or the user can choose to go full screen with the image.  On the full screen, the user can download the image to your computer (by doing a "Save image as" in your browser).  The household members are listed below the image.

Finally, the user is told that if this match is correct to extract the information by the record by editing the person profile in the MyHeritage family tree.  The user is given the choice to Reject or Confirm the match.

5)  After this first look, I have these thoughts:

*  The Record Match system works pretty well as far as it goes.  It is limited by the record collections available to MyHeritage, including free collections like Find-A-Grave and Ellis Island, and subscription collections, like the census and newspaper collection.

*  It is unclear if the Record Match number covers EVERYONE in my 40,000 person family tree.  If it doesn't, then I wonder if the system looking for more matches automatically, or do I have to instigate that somehow using the Research menu?

*  There is no indication of the limitations of each record collection.  I know that I have many more persons in my tree that are in the 1940 U.S. Census - the MyHeritage matches for my people are in Connecticut, New York, and Pennsylvania.

*  We all know that many records were created with errors in names, locations, ages, etc.  It is unclear to me how the Record Matching technology decides on the range of the variables.  Are names coded using Soundex or similar?  Do ages have some year range?  The MyHeritage SuperSearch matching in the Research tab permits these variations.

*  The Filtering options appear to be logical and useful.

*  The MyHeritage Record Search technology will help me find additional information on persons in my family tree.  While I have mined databases like the Social Security Death Index and Find-A-Grave extensively, I have not done extensive mining in the Newspaper Archive collection.  This will be my main concentration in the weeks ahead.

*  I believe that this is the only record matching technology that looks in newspaper records without  entering names into a search field.

*  The MyHeritage Record Search is an excellent advancement for MyHeritage users as long as they have a subscription to the record collection providers.  This is a first step for MyHeritage because the number and type of record collections available are limited.  If they can add access to many more record collections (e.g., U.S. Census, England Census, military records, passenger list records, and more), then they may rival Ancestry.com in the future.

6)  Disclosure:  I have a complimentary subscription to both MyHeritage.com and WorldVitalRecords.com courtesy of MyHeritage, for which I am grateful.  However, this does not influence my objective opinions in reviews of these websites and their products.

The URL for this post is:  http://www.geneamusings.com/2012/09/first-look-at-record-matches-on.html

Copyright (c) 2012, Randall J. Seaver.

6 comments:

Kay Haden said...

Randy,
I have a paid subscription to MyHeritage and to WorldVitalRecords through MyHeritage.
I checked out the Record Matching today and what I discovered is that they attempt to charge me credits for Matching to FindAGrave! At 100 credits per record just as though it was behind a pay wall. Of course I can use the URL provided without paying.
I also did match a record to one of my persons and nothing really happened. I already had the correct data but just thought it might be worth the 100 credits to find out what happens. Nothing appears on that persons profile to reflect the match to the source of that record match. At this point I don't understand the value!

Gilad said...

Thanks Randy for this quick look.
Hopefully you will be able to take a deeper look soon.

A few comments:

1. The default filtering of one star means you will see matches with a confidence of one star or higher. Matches with less will be hidden in the interface, i.e. the weakest matches are hidden by default. Users with few matches who who are desperate for any leads can change that setting by lowering it, but things work well without changing the default.

2. Yes, ALL 40,000 people in your family tree were covered.

3. The indexing of the 1940 US census is still ongoing at MyHeritage. We are putting an emphasis on the quality of the indexing, not on getting it done quickly. So for now you will get Record Matches for some of the USA states, including New York and the others you mentioned. As we add more states you will receive new Record Matches automatically until the 1940 US census is fully indexed.

4. It is complex to explain how Record Matching technology decides which variables to take into account and when, it would take a long white paper to do so and the algorithms are proprietary. It would be great if you could comment later on the correctness of the matches that were found for you from your point of view as a genealogist, and the new information they allowed you to add to your tree, especially regarding newspaper matches which are unique.
I can mention that matches are indeed flexible with names and synonyms, and various types of soundex are supported in matching, and information like year discrepancy is allowed, depending on other circumstances.

5. MyHeritage users do NOT need a subscription to any record collection providers. Only to MyHeritage itself (Data Subscription) or to have pay-as-you-go data credits on MyHeritage. MyHeritage users with a World Vital Records subscription receive a complimentary MyHeritage Data Subscription automatically (since we now own World Vital Records) so they not need buy anything else to enjoy Record Matching. For users who do not want to pay at all we provide free matches with all collections (as an extract) and full lookups of Find-a-Grave and Ellis Island matches.

6. You can count on us adding more USA census records soon and the UK census records, and other types of records. This is a focus for us. Our aspiration is not to be a rival to Ancestry but to provide a superior product and experience for the users. This may take us more time, but this time last year we didn't have any historical records at all, nor any search engine, nor Record Matching technology, so imagine what we'll have one year from now. SuperSearch and Record Matching work in 38 languages, not just English and 4 others. So users around the world who are not English speakers already have products they can use on MyHeritage. And MyHeritage provides unique content so that Ancestry users will find value in receiving Record Matches on MyHeritage, even if they already have a subscription on Ancestry. Many genealogists would appreciate additional information and new records and discoveries, especially when found automatically. also This comes with Smart Matches which also cannot be found on Ancestry. We hope that users will eventually want to make the switch (noting that a MyHeritage Data Subscription costs $76/year, less than a third of an Ancestry subscription which covers all the records).

An answer for Kay: the value in the Record Matches is in the matches found automatically for you. What you tried to do was not to "match" but to "confirm" a match which is to indicate that it is correct. The value in confirming is for marking matches from which you have extracted information, as you go through the matches and process all of them. Otherwise when having many matches you won't be able to easily remember which ones you've already processed. Find-a-Grave matches are free and we will apply a fix so that confirming them will be free too and will not require any credits.

Best regards,
MyHeritage\Gilad

Pat and John Walker said...

I also had a current subscription to WVR, as well as my MyHeritage Premium Plus subscription and at first, they were trying to charge me for access to the actual record image and results. I checked under My Account, My Purchases and it showed "No Data Subscriptions"

I had to email them twice to get them to acknowledge my subscription to WVR. I wonder how many WVR subscribers did not understand that and went ahead and paid "double" for the access.

Esther Shuman said...

Hi Pat,

We have looked into your situation and since written you an email back.

Thank you for your feedback.

Esther / MyHeritage Team

John Beardsley said...

I have a MyHeritage Premium Plus account and a tree with over 40,000 names. I was talked into upgrading to the Premium Plus account because of the Record Matching feature that was described by a salesman.

Months later I have ZERO record matches. It seems I was sold a bill of worthless goods.

Is there a method to apply for a refund?

Pat and John Walker said...

John,

I wonder if there is a setting that needs to be changed. Also, I have a some Premium Sites (vs Premium Plus) and they also get Record Matches. If they cannot help you, you can ask for a pro-rated refund, or you could get a refund to downgrade to Premium and still get Record Matches--once they figure out what is wrong with the feature on your site.

You ought to ask for a free subscription, since the features they advertised and you paid for are not functioning.