Continuing problems with Possible Duplicates suggestions

LegacyUser

Adrian Bruce said: Anything is possible but I suspect that idiot merges are more likely to be the result of people blindly following hints than anti LDS vandals, who are more likely to just edit garbage into the various items than to find a person of a similar name to merge in.

Of course, they would say that, wouldn't they? (Cynical and/or confused grin).

LegacyUser

Ryan Torchia said: Really. Really? You have any idea how many James Wrights who married a Mary a couple hundred years ago there are? How many John Mullinixes were born in England in the early 18th century? And how many ways there are to butcher the name "Mullinix"? And how many lazy census takers just wrote "Mullins" in North Carolina and Missouri a hundred years later? Well, I have a vague idea.

You ever create a profile for somebody who lived a couple days and immediately had to dismiss the thirty suggestion FS generates because their parents used the name for three other children, the last of whom survived and had fifteen kids of their own? And oh yeah, they also have ten cousins with the same name? I have.

You can take the first names Michele, Vincenzo, Calogero, Accursio, Antonio, Antonino, and Giuseppe, and the Surnames Sclafani,, Montalbano, Friscia, Fauci, Maniscalco, Sabella, Chiarello, and DiMino, and in the town of Sciacca, Italy in the 19th century, you will find ten to fifty people for every combination of those,usually married to a woman with one of those same combinations except with an 'a' at the end of their first name, many of whom appear in thousands of records by just their first name, and most of whom had their names absolutely mangled after they arrived in New York City.

So please don't think for a second that you're the only person who deals with that. I'm fine with dismissing a few hundred suggestion to have a system smart enough to suggest that the "Tanazia Pizzo" mentioned in his child's New York marriage record with his wife "Paula Catanjaro" might actually be the Ignazio Piazza who married Paola Catanzaro 3000 miles away. I want the system to know that a birth or christening location might just be a place where the person lived later, and not assume every piece of data in the record is flawless. My tree would be a lot emptier if it did.

LegacyUser

Ryan Torchia said: These sources first got attached through a few bad merges from May 10, 2019. The person who did that appears to have never come back to the profile since.

Those merges were probably suggested because the profiles that got merged were essentially the same names, "John Knight" and "Hariott", with no other vital information on the record itself (i.e. no conflicts in the vital information data between the two profiles). England was only mentioned in the attached sources and their childrens' profiles. The system doesn't pick up on this, and I don't think it's reasonable to exclude suggesting the match even if it did.

The person merging should have checked vital info for the children and looked at the sources being merged and noticed it. That's careless, but it's not malice. They didn't actually think one person lived in multiple places at once.

Reverting the merges would have been better than just disconnecting.

LegacyUser

Paul said: It appears we will have to "agree to disagree" on this one, Ryan. Oh, and no, that fact I raised this issue is partly due to the fact that I DO realise there are many other users who have the same issues as me. Certainly I realise many others, including you, are spending lots of time trying to get things right.

However, I note you do not address the serious issue of the users (who surely you have come across, too) who have effectively wiped out of existence many deceased individuals. Using my James Young example, several persons of that name would no longer have existed in Family Tree had I not unmerged them all. The idea of this project is to have one ID for as many individuals as possible who have ever existed on Earth - not to leave thousands of individuals without any ID, through incorrect merges that have resulted from implausible "possible duplicate" suggestions being placed on person pages.

LegacyUser

David Newton said: Oh it's most likely carelessness alright. Grossly incompetent carelessness.

As for reverting the merges, rather than doing what I did? Why? Ordinances? Can't see them, have nothing to do with them, not my problem. Besides tracking down those merges in the changelog and unmerging will not produce helpful results. New residences and christenings resulting from the merge will not be reversed. Sources will still be left in a mess.

You see the only reason I use FSFT is for the genealogy. I don't do it for religious reasons. Vicariously I have picked up a great deal of information about Mormon doctrine and practices, but the most that is for me is mildly interesting information. I care about the information being genealogically correct. What I have done to that profile leaves it in a vastly more genealogically correct state. Perfect? Certainly not. If a Mormon wants to go in and sort out the ordinances then they're more than welcome to. That's their business, not mine.

LegacyUser

Paul said: As I spend the evening detaching sources from the incorrect Wrightson family to whom a careless user has attached them, they are immediately reappearing on the same person pages as Record Hints. Now I am having to dismiss them as "Not a match"! Both Possible Duplicates and incorrect sources are involved.

No wonder people want to give up on Family Tree. A complete nightmare over the last week. Hours of work and none of it "productive". Shame there was nothing to watch on the TV!

LegacyUser

Ryan Torchia said: Because if you unmerge them and make a minimal effort to distinguish them they won't be suggested as merge candidates again, and people are less likely to merge them if they're easily distinguishable. Those individuals are just as much victims of these careless edits as the people you care about. You leave the tree itself in better overall genealogically accurate state if you clean up both sides of a bad merge.

Unmerging can also be easier when it was two individuals with full families than trying to cover up the result of a bad merge. Plus, the merge can be tracked down through the edit history for related people and unmerged later, which will revert much of the work done to clean up the records anyway.

I'm not an LDS member either, and I don't have any personal interest in ordinances, but still, this is their site which they are allowing us to use for free, and they're putting in a ton of effort into preserving something very important. It seems like some basic respect is deserved.

LegacyUser

Ryan Torchia said: Absolutely, there are definitely solutions that don't involve limiting correct suggestions at the expense of incorrect ones. I would totally support a tiered system like that, or a per-user setting for how strict or loose a match can be, maybe based on experience and how many edits have been completed. Or blocking merges for new and inexperienced editors. Or even the ability to lock down parts of profiles that we have a high level of confidence in so they don't get overwritten.

I don't know if there's any way to even report bad edits in a way that gets tracked so that repeat bad offenders get stopped. And often the damage I'm undoing is months or years in the past.

I still think the biggest problem is info from bad merges and mis-attached sources that get propagated out to GEDcoms or third-party apps, and then get endlessly re-imported.

LegacyUser

Adrian Bruce said: "if you unmerge them and make a minimal effort to distinguish them they won't be suggested as merge candidates again, and people are less likely to merge them if they're easily distinguishable"

My tactic if they can be separated somehow, either by unmerging the profiles, restoring the deleted-by-merge profile or plain simple recreating the merged person from new if all else fails, is to

(a) see if a merge is suggested again automatically - in which case I explicitly mark them as not-a-match - or if not...

(b) attempt to re-merge the two profiles manually by PID then, at the point when the system asks you what to do, then I say, "Not a match because...".

Either way, I like to have a "Not a Match" marked up. Though I can't swear to always do that.

LegacyUser

Adrian Bruce said: Endless repeats of the same misidentifications are definitely an issue. I have a brick wall ancestor who I spent ages on recreating in FSFT because she'd been merged with someone 2 counties away back in pre-FS FamilyTree days (so no possibility of unmerging or restoring deleted-by-merge profiles.).

Gradually the crud data is creeping back into her profile - I think because the crud data is in one of the FS genealogy files (the fixed ones) and people find it, assume it must be true and copy it over - sources? Isn't this guy's tree enough?

LegacyUser

Ryan Torchia said: Same here. Since the system can be laggy, if I create a family from scratch to link a census report just to avoid it getting mistakenly attached to the wrong family, I'll force the Not a Match as well just as precaution.