Algorithm Updates

roberthparker3

🆕 Update: 10-3-24 algorithm improvements:

In case you missed it, here are some recent improvements to our algorithms:

• Better checking of patronymic names
• Better handling of name consistency with first and middle initials
• Reduced flagging of some data problems when supported by sources

Thanks for continuing to share your feedback!

Update: 5-28-24 recent algorithm changes:

• Smarter name comparison for sources with only a surname
• Added geo, jas, jos, and thos to nickname list
• No longer calling out mismatch between Mc Donald and McDonald.
• Started checking for issues where a parent dies too young to have children
• Started checking for issues where a person dies too young to be in a couple relationship
• Started checking for parents getting married multiple years after the birth of a child
• No longer expect a death source tagged if also missing death date and death place
• We are now scoring persons born in South America between 1800 and 1920

Update: 4-29-24 More changes to the algorithm based on your feedback:

• No longer flagging missing burials if a cremation fact is present
• Only call out marriages at an older age if it is the person's first marriage
• More leniency when comparing dates on census records
• Less leniency when comparing dates on other records
• Name comparison issues now return the name in Proper Case

Thank you for your valuable feedback!

Update: 4-17-24 Thanks to some of your feedback, we have released some additional changes:

• We have improved our city detection algorithm (we have heard your complaints about New York City, New York, United States)
• Requiring a christening has been changed. Giving credit for having a christening will improve the score but there is no punishment if one is missing.
• We currently have a nickname list and are accepting Wm for William.
• The bug about dismissing issues is now fixed.
• A more lenient algorithm for place consistency is now in effect. This should help with places on a time boundary.

Update: 4-11-24 We have recently released some additional algorithm updates:

• We now require two tagged sources for a name and a birth conclusion.
• We have adjusted the scoring algorithm to more heavily weight the "Data is conflict-free" category this will result in lower overall scores than you may be used to
• We are now scoring persons born in continental Europe between 1800 and 1920

As always, thank you so much for your continued support and Feedback!

4-5-24 Update: We have recently released some algorithm updates, many based on your feedback:

• Give credit for alternate names when comparing names with sources
• Create an issue for each source that can't find a matching name
• Score impossible event order issues more harshly
• Stop calling out untagged marriages
• Allow tree persons from Newfoundland to be scored
• Fix bug comparing residences with tagged sources from 1940 U.S. census
• Sort some birth issues by the age of child

We look forward to hearing more about your experience with the Profile Quality Score.

3-27-24 Update: Today the Profile Quality Score started checking for some new quality issues:

• Biological children born too close together
• Too many biological fathers
• Too many biological mothers
• Unlikely migration patterns

If you happen to stumble upon any of these, we are interested in your feedback.

We also officially retired the standalone Prototype 1.0 (which had a limited release in October 2023)

Áine Ní Donnghaile

Robert -

Thanks! That's quite a list of fixes!