Monday, June 11, 2012

MyHeritage Announces Record Matching Feature

I just finished watching the MyHeritage press conference (on a webinar - watch it at with Gilad Japhet.  Dick Eastman has a copy of the "MyHeritage Hits One Billion Profiles" press release here.

During the conference, MyHeritage announced a new feature called Record Matching.  Here is a screen shot of the slide with details:

The slide notes:

*  This technology will be released in July 2012 to all MyHeritage users.

*  Record Matching intelligently and automatically compares the 1 billion family tree profiles on MyHeritage to the 4+ billion historical records to find matches.

*  The site will translate names between languages and apply synonyms to find record matches.

*  This is the first record matching service to find matches in newspapers, yearbooks and other true text documents.

*  Record Matching has 98% accuracy and an average of 15 matches per 100 people. 

I was first shown the idea for this technology in February by Gilad while we were at RootsTech.  He asked if they could use my tree on MyHeritage to find record matches to demonstrate the technology.  I said "YES!!!" because I saw the potential.  Last month, Gilad emailed me telling me some of the matches they had found, and he asked me if I would consent to some of my matches being used in the press conference presentation today.  Again, I said "Yes" being the semi-famous yet unassuming genealogy publicity seeker that I am.

The first match was for my first cousin, twice removed, Frederick Thomas Blanchard, a university professor.  The Record Match system found his wedding announcement in an Oakland (CA) newspaper:

The system also found a college yearbook summary of his achievements:

Gilad used two other examples for persons in my database, and several examples from his own tree.

In Gilad's recent email to me, he noted that there are 3,925 Record Matches for persons in my MyHeritage tree, and they include 825 newsletter articles and over 1,700 cemetery records.  I have a lot of work to do!

I asked the question (via chat):  "How will the Record Matching be delivered to a user?"  The response was via a weekly email with links to the matches in the record collections, and by links on the MyHeritage tree (website or on mobile).  Users will be able to filter the results by criteria and sort them by confidence level.

While it wasn't explicitly mentioned, I think that only MyHeritage subscribers will be able to use the Record Matching feature, and only subscribers to the site will be able to access the historical records (including the newspapers) on that site.

I do want to correct Gilad's description of my database as "... a meticulously documented tree ... with a source for each and every fact."  That isn't exactly true... I do have many sourced facts for many of my ancestral families, but not as many for the collateral lines in my 40,000 person tree.

I look forward to trying out the Record Matching feature when it becomes available on MyHeritage.

My thanks to Gilad Japhet for using my family tree for some of his examples.

The URL for this post is:

Copyright (c) 2012, Randall J. Seaver

No comments: