HUGE number of incorrect Record Hints on Scandinavians in the tree

LegacyUser

Dale Hein said: I wish FamilySearch would stop putting up all these Record Hints every time another Swedish film is indexed. 90% of the hints on my ancestors' pages are wrong. Most have different wives, lived in different places and had different kids. Swedish people didn't have that many different names. Just because someone is named Jonas Mansson doesn't mean he is the Jonas Mansson in some other town with a different wife who that Jonas Mansson married 10 years earlier or later. This is so frustrating, because if I don't stay on top of all the Record Hints as they are added, most of which have to be declared Not a Match, somebody who doesn't know what they are doing attaches them and adds wrong wives and sometimes merges them, adds wrong children, etc. and messes up families I found by pouring over microfilms page by page for many hours for over twenty years. And then somebody else is going to come along and grab a green temple and off they go to seal the wrong wife and child to my ancestor. Because if FamilySearch puts a Record Hint there it must be true. Please, please, please stop putting Record Hints on Swedes and Danes if the system can't do a better job than it does. This is serious business, not a game. Please.

LegacyUser

Paul said: I think someone must have messed around with the algorithm again. I was really pleased to see things appeared to have been tightened-up, especially for hints relating to English relatives. But recently the situation has become bad again and if a person had the same name and same (or even similar) parents names, a hint appears regardless of the part of the country or even for people who lived in another century.

There used to be a thread on which these could be reported, but as things had improved so much I have lost track of it. Robert (Kehrer) - hope you are still following posts in this forum. If so, please get your engineers to take another look at this problem - it can take too much of our valuable time to dismiss all these silly hints.

LegacyUser

Brian Jensen said: Dale, Paul,
Can you please share with us specific examples (i.e. Person Ids) of these incorrect hints?. In December we made several improvements to our Hinting algorithms but we are always looking for more ways to improve the quality of our Hints.

LegacyUser

Paul said: Brian

Thanks. I have dismissed ones found recently but will post here when I find some more. However, you can hardly call it an improvement when the hints are still sometimes for the wrong century, and/or are presented even when the correct event has already been attached.

LegacyUser

Dale Hein said: Brian, I have a list for you with links. Can I have your email address? Or you can find mine easily enough if you don't want to post it here. My username is dlhein. The sooner the better, because I just found 15 record hints on an ancestor whose been in the records since 1914 and half of them are for a totally different family, and I don't want to fix it until you see it, but I'm going to have an anxiety attack worrying about. Thanks.

LegacyUser

Juli said: Dale, wouldn't it be easiest to just post the PID for the ancestor with those 15 record hints?

LegacyUser

Dale Hein said: Sure. Here is that one, but I have a lot more than one along with explanations and I didn't want to put a huge post. https://www.familysearch.org/tree/per...

LegacyUser

Dale Hein said: Like that Anders Nielsen was only married once. And even if he wasn't it's obvious he couldn't have been having two families with two women in two different places with children being born at the same time. And I wanted to remind him that there were probably thousands of Anders Nielsen's in Denmark in the mid 1800's, etc.

LegacyUser

Dale Hein said: In SLC maybe, but not likely in Denmark. But if I leave that there, I can guarantee you somebody is just going to attach all of those hints without even looking at the data. It happens ALL the time. And when they combine families and then merge members of two different families and do all other kinds of damage all based on a totally crazy Record Hint the system puts there, and it ruins all my years of research it's very upsetting.

LegacyUser

Brian Jensen said: Dale, thanks for sharing this example and giving me a chance to review these Hints. Of the 13 Record Hints I believe that 6 are perfect matches, 2 are possible matches and 5 look like good Hints on the second woman/extra child. The second woman was added in Sept 2017 with this note: "Mads Christian Andersen was the illegitimate son of Jytte Dorthea Madsdatter and Anders Nielsen. Anders and Jytte never married." You may want to reach out to the individual that added this relationship. Our Hinting algorithms can only use the data that's provided in the Persons / Records

LegacyUser

iLoveMyLife02 said: What about this set of Possible Duplicates?
https://www.familysearch.org/tree/mer...
We have a marriage date of 1706 and a DoB for Marrie as 1629 (though, admittedly, with no sources attached for that particular fact, but she had a child with Hendrik Jansen 9NH3-HNY that was born in 1648). I know that people do sometimes get married at age 76, but this just doesn't seem very likely.

LegacyUser

Dale Hein said: Brian, I knew some of them were right. And I do always send a message to those who attach wrong records. Even if that child was illegitimate, Jytte had more children with an Anders Nielsen. Probably the same one, but like I said there were probably thousands of Anders Nielsen's. Trust me. I know how to do it right and I will, but if your algorithms can only use the data that's provided in the Persons/Records, then it shouldn't match an Anders Nielsen in Denmark who is already married to someone with a totally different name and living in a different place with another Anders Nielsen who isn't. Especially in Denmark or Sweden. My goodness! I have more examples and I'll keep collecting them. And it isn't only on the Scandinavian ancestors. I find them on all nationalities.

LegacyUser

Dale Hein said: Brian, Here are some I already fixed. And there's been a lot more (not only Swedish) but I don't have time to look for all of them. I need to keep looking for new Record Hint problems and problems caused by users uploading GEDCOM files.

https://www.familysearch.org/tree/per...
https://www.familysearch.org/tree/per...
https://www.familysearch.org/tree/per...
https://www.familysearch.org/tree/per...
https://www.familysearch.org/tree/per...
https://www.familysearch.org/tree/per...

LegacyUser

Dale Hein said: Can't you at least put a big red warning message at the top of the Record Hints saying something like "CAUTION! Not all Record Hints will be for the individual whose page they are on. Please check all data carefully and click Not a Match with a clear reason statement if any Hints are not for your ancestor."? Or make it pop up every time users click Attach?

LegacyUser

Dale Hein said: Here's another one: https://www.familysearch.org/tree/per...

And his wife's last name isn't Bech. In fact, her first name isn't Karen either. It's Maren. And nobody -- not him, his wife, any of his children, his parents, or any of his siblings have Vejle, Denmark on any of their pages. Vejlby Parish in Aarhus County, yes, but not Vejle County. And it isn't on his wife's parents or siblings either.

But I did a find by name for Karen Hansen Bech, and there are 12 results with that name or some form of it. And the first two are exactly Karen Hansen Bech and the husband on both is exactly Niels Thomassen, not Niels Thomsen. So why did those two Record Hints find their way to my Niels Thomsen and Maren Hansen's page and not Karen Hansen Bech and Niels Thomassen's page, the names on the Record Hints???

LegacyUser

Dale Hein said: And what do you know? The very first result had a son Thomas Nielsen born and died in 1860! Now I really wonder why the hint didn't go there instead. Ugh. I just clicked Review and Attach and it's already attached to the other three. It says Detach instead of Compare, so I clicked to see the ID's and it's them! Ugh.

LegacyUser

Dale Hein said: Brian, I just figured out that my Anders Nielsen's page is all messed up because of either an incorrect merge or combine in nFS. Those Hints are NOT for him. And that child is not an illegitimate child of my Anders Nielsen. And notice the parents are missing from a 1922 SP for my Anders Nielsen. I see an incorrect merge. This might take awhile to fix, and I will, but those Record Hints do not belong on this Anders Nielsen's page.

LegacyUser

Dale Hein said: Oh, and I forgot-- I did reach out to the woman who added that note. We are distant cousins who have been FT friends for a couple of years. She is an awesome researcher, so I was surprised to see it was her that added that note, and that is the only reason I haven't gotten on this mess sooner. I didn't want to start until I hear back from her.

LegacyUser

Dale Hein said: Two more obviously wrong Record Hints:
https://www.familysearch.org/tree/per.... Totally different county, mother's first name is different and and both children born 2 and 5 years before parents were married.

I'm going to hurry and dismiss them right now, so you'll have to look in the Changes to see them.

LegacyUser

Paul said: Brian

Here's a good example of the problem with individuals from England. In spite of my having inputted an Essex christening and a marriage to Mary MARTIN (in Essex) I am still presented with this possible duplicate! (I am not quibbling about the difference in surname as I have John HAYWARD as an alternate name on his profile.)

LegacyUser

Paul said: Brian

Here's an example of how both the record hint AND possible duplicate algorithms are not working well for those of us with English ancestry.

In the case of the record hint, this is presented for an event in NORFOLK, in spite of all events attached to the immediate family relating to the county of ESSEX.

The possible duplicate suggestion shows a DORSET birth & marriage for John (and a spouse named Mary LEAGUE), in spite of my having inputted an ESSEX christening & marriage and a spouse named Mary MARTIN.

(The difference in surname does not concern me - especially as I have inputted John HAYWARD as an alternate name for John HOWARD.)

LegacyUser

Dale Hein said: Thank you to those who are adding examples! I hope the engineers realize that it isn't that we are complaining because we have to dismiss incorrect Hints. It's that we're worried about the people who just assume every Hint they see is for the ancestor whose page it's on and they add it and by doing so are saying bring on more incorrect Hints. And they are adding incorrect relationships and data to our ancestors and causing incorrect temple sealings.

LegacyUser

Paul said: Exactly, Dale! I know from previous experience that an inexperienced user could well have added the incorrect hint AND completed a merge with a totally different individual, because they foolishly trust FamilySearch are presenting them with reliable suggestions.

LegacyUser

David Newton said: Yep. False positives like those that are extremely obvious to an experienced genealogist would likely be added willy-nilly by someone who is not being careful enough or who hasn't taken the time to check the geography properly.

The question is why are they tripping the algorithm? What factor is the algorithm giving too much weight to and/or what factor is the algorithm giving too little weight to. One possibility is that the algorithm is giving too little weight to surname rarity. For John Smith just popping up random cases of John Smith would be useless, but for Extremely Unusual-Surname popping up random occurrences would actually be useful as the name is so rare.

LegacyUser

Dale Hein said: Not knowing anything about computer programing at all, I did have a thought. When you do a Search Records from a person page or when you just Search > Records from the Search tab at the top, you get a huge list of results some of which look like they might be the person and a bunch that probably aren't. And when you are really doing research that's a good thing, because you know they can't all be right, but you have a lot to consider.

Is there any possibility that when the system automatically adds Record Hints after a film's indexing is complete that it is just as liberal with where the Hints go? Because then it is NOT a good thing, for the reasons everybody has been saying -- people who are NOT doing research will add anything FS says goes with that person because FS must know, and they really want to help out. ha ha.

So if I am even close to understanding the way it works, can't you just only let a Record Hint go on a page if 3 things match -- name, date and place. At least the name being very close, the EXACT year and the EXACT state or county. If that means a lot less Record Hints, that's good! Let the inexperienced attach the obviously correct and those who love and are good at real research will find the others. It isn't that hard once a film is indexed. It isn't even that hard once a film is just digitized.

LegacyUser

Juli said: One factor that the algorithm definitely doesn't give enough weight to is the wife's surname. I have had hints where the wife's name in the suggested record shared zero letters with the actual wife's name. The algorithm also gives insufficient weight to locations; I swear, sometimes it figures "eh, same continent, gotta serve it up."

LegacyUser

Tom Huber said: As one completes the record for a relative, new hints can appear because of the connections that can be made.

But... and I've found this to be the case far too many times... if another user attaches a source from a hint they believe is for the person (and it isn't), additional hints can also appear that are not for that person.

After detaching a hint that has been erroneously attached, come back after an hour or two or the next day to mark the same hint (it reappears) as Not a Match.

LegacyUser

Dale Hein said: Another really bad one. https://www.familysearch.org/tree/per...

The name is VERY different. I guess it ended up here because her husband was Ernst Hein and they had a son William. But none of my people with those names ever lived in Ohio. And there is already a wife and marriage record for CARL William Hein. He married in Germany and never lived in Ohio and all of that is in the tree, including his real marriage record.

LegacyUser

Dale Hein said: Another one: https://www.familysearch.org/tree/per...

This one also has a possible duplicate that has so little information nobody could say if it's a duplicate, but I know it isn't because the name is wrong, he was never married to anyone by that name or had a child by that name and his family lived in a different parish for at least 4 generations.

And the two Record Hints are wrong, too. Ugh. I think it's getting worse. I am find more wrong than right.

LegacyUser

Dale Hein said: By the way, I'm dismissing all of the things I find so if anybody wants to see them they'll have to check the changes. If I don't get them off of there I know somebody will merge the duplicate and add the hints.

LegacyUser

Dale Hein said: Another one: https://www.familysearch.org/tree/per...
These last ones have all been Germany or US, so it isn't just Scandinavian records. It's all of them. It's a full-time job trying to save my tree and all the research I painstakingly did over the years.

This one is really bad --9 hints all wrong. 8 are for Nicholas Kuns other wife Catharina Phillippi, not this wife Catharina Preis! And the 9th one is for a totally different woman who has nothing to do with this family. And I can't fix it right now. I have to go to the dentist. That's scary!