Person Page/Research Help: Tweak the Research Help algorithm to omit unreasonable suggestions.
Edited for formatting by a moderator:
- My problem is ... I get research help suggestions that aren’t logical based on the dates of the people involved.
- When I ... review suggested research helps,
- I want to ... see suggestions that are possible based on my ancestor’s life information.
- Because ... I don’t want to spend time dismissing research helps that aren’t plausible.
Example w/screenshot.
Original Post:
The earlier born child shown here has a death date shown / inputted, so if should not be necessary to suggest she might be the same individual as a later born child given the same name.
Also, in this case, the suggestion of a missing child is made, yet his first wife died only a year after the marriage.
I'm sure I've encountered many other examples (also found when adding a child of the same name to parents), so surely "death" should be a factor here?
Comments
-
I had a similar one recently - a couple had 2 sons born within 11 months. Both were named for their father and both died within 3 days of their respective births. Both had birth and burial records attached. The hinting algorithm suggested I should merge them as duplicates. Nope.
2 -
Sorry my screenshot has not come through. I first posted this on 18 March and it has been subject to delay in publication as I did not stick to the new guidelines regarding the format of "ideas" we should now use when making a suggestion in this section. Unfortunately, I only took a screenshot on this occasion (i.e., I did not add the relating URL) and did not save a copy.
Hopefully, I will be able to retrace the Details page in question. Meanwhile, thanks to @Áine Ní Donnghaile for confirming other users are experiencing these unwanted hints.
(BTW - I have copied this post - written 26/03/2024 11:07 - as I expect it, too, will disappear for a time whilst subject to moderation)
1 -
The problem is, sometimes such profiles really are duplicates. For example, I just read a report written by Elizabeth Shown Mills (first one on this page: https://www.historicpathways.com/researchreports.html) on a Louisiana family where no two records ever agreed about anyone's age or birthdate -- but this was all perfectly normal for that context.
Even in contexts where precise birthdates and birthplaces are available, date discrepancies of several years do not necessarily preclude the profiles being for the same person. I see this most often in cases of "same person, different ends": one profile is based on the person's birth and baptism, while the other is based on her marriage and the births/baptisms of her children. Marriage records often record just ages (if you're lucky), not birthdates; even if the recorded age is correct, the calculated birth year may be off by one, creating a situation that looks exactly the same as when a child was named after a deceased sibling.
I agree fully that needing to dismiss impossible hints is aggravating at best. I'm especially annoyed when the hint ignores women's surnames completely, and when the places don't match on anything beyond the country. It makes me wonder: how many unrelated families have been combined, to "teach" the algorithm to make such hints? In fact, I think one reason we object so strongly to bad hints is that they're scary: we've all had run-ins with inexperienced or ill-informed contributors who took the computer's word on everything. What if one of those people got to the impossible hint first? But I don't know if the algorithm can be taught to tell the guesses from the well-supported facts.
I suppose if it were easy, we'd all be out of a hobby, because the computer could do it all for us. :-)
3 -
In case it is useful, here's the URL of the family where I encountered the issue: https://www.familysearch.org/tree/person/details/9WY6-8MN The 2 sons August were born 12 Jan 1842 and 17 Dec 1842.
The dismissed help - a hint that the 2 sons are duplicates: https://www.familysearch.org/tree/person/research-help/MQ16-6XF
0 -
Still, it would be useful to be able to force recognition of a death date. I've had hints of people born several decades after the death date which makes me think it is not even being considered.
0
