Issue with an index to "England and Wales, Census, 1911"
The issue is with my distant relative, John Lofkin, and his "England and Wales, Census, 1911" index on https://www.familysearch.org/ark:/61903/1:1:XW3Z-BCY
According to that index, these are the summary details of John's spouses and children:
If, however, you look at the image, John Lofkin is the first line in the household followed by his housekeeper, Ellen McManus. The children, Fred thru Elsie are all Lofkin on the image. Not McManus as above. William McManus genuinely is William McManus.
So the first reaction is that this is an indexing error and McManus has been propagated down from line 2. Except… The index record explains that this is a "database and images" from FindMyPast. So this suggests the indexing error comes from FindMyPast (FMP).
Except… The index is (currently) correct on FMP - the children, Fred thru Elsie are all Lofkin on the FMP index. Not McManus as in FS.
Is it possible to explain where the error of McManus for the Lofkin children comes from?
If the FMP index has always had the correct value of Lofkin, then this suggests a corruption somewhere in the process of sending the index to FS and loading it there. Perhaps (wild guess) because the housekeeper is on line 2 and the "son", "daughter", etc, have been taken to apply to her rather than the head of household?
On the other hand, if the FMP index originally had the incorrect value of McManus then maybe someone like me requested a correction in FMP (to Lofkin). In which case, when will the corrected FMP indexes get thru to FS?
Is it possible to work out what the original values for the index in FS were?
Answers
-
I believe the closest you might come in getting an answer to this is by accepting your own suggestion that the FMP indexed record was originally the same as it remains in FamilySearch, but was changed following a report from a Find My Past subscriber.
As you know, we rarely get any "official" responses to such questions. The mystery still remains as to why millions of 1871 census records disappeared for a considerable period from FamilySearch, then just as suddenly reappeared. If the FS record does eventually match the current FMP version, it will probably be as part of one of the periodic updates FMP passes to FS. Keep an eye for the next update - by monitoring the details at https://www.familysearch.org/search/collection/list?ec=region%3AEngland%2CplaceId%3A1986340%2Cregion%3AUnited+Kingdom+and+Ireland&fcs=placeId%3A1986340&count=20
I don't know why the 1881 E&W census collection does not appear on the list, but I see all other collections, barring the 1911 one, were last updated in June 2024. The 1911 census appears not to have been updated since August 2019, however, so perhaps your issue will be resolved at the next update from FMP - which will hopefully be quite soon!
0 -
Regarding my mention of the 1881 census issue, please see my post at
where @Gordon Collett had helpfully provided the "workaround", which enables the 1881 collection to appear in the list of available E&W census collections.
1 -
Hmm - the personas on the source index read "McManus" (i.e. the wrong name) back in July 2022 when they were attached to the profiles, @Paul W - took me a while to find evidence of that, but the change history of the link between profile and persona in the source index, shows it.
But whether it's always been wrong in FS (probably), and whether it's ever been wrong in FMP (unknown), I don't know. (I don't see any way of deducing the history of the source index in FMP)
I can say that it's actually unlikely that I would have picked up on any error in FMP if if if such an error existed because by the time I did this family on my PC, the 1911 was on Ancestry, which is my usual census source, so I used that.
0 -
@Adrian Bruce1 FWIW - I've found FMP Support very responsive when I've written with a specific question. Staff there may be able to tell you if there is a change history on that page/piece.
I wrote to FMP Support about a broad error in the RC records of a County Sligo parish. The parents' surnames had been badly cross-threaded, making the register almost useless. After our exchange, the entire parish was re-done. Their promised follow-up (We'll let you know how we get on) didn't happen, but the records were corrected.
1 -
Áine - worth a go, so I've just mailed FMP to see if they do have a change history for that entry. I emphasised that their index is correct!
1 -
FMP responded to my query thus:
"… We have no recent evidence of that record being updated through customer corrections. That's not to say that it hasn't changed at our end just that we couldn't say for sure how or when (if at all). …"
(My emphasises)
So since the (incorrect) McManus names were on the index in FS in 2022, it could be that the change was updated in FMP after the 2022 data was sent from FMP to FS but that isn't in the "recent" log. (I have a feeling FMP changed stuff in relation to updates in the not too distant past).
So no real affirmation of whether anything changed in FMP or whether nothing changed in FMP and it's a software error in FS. Which last possibility just feels a bit odd to me.
1 -
Yes, I can confirm that FMP changed their updating process - I think in early 2024 or late 2023. Emails received from FMP acknowledging a submitted correction started going to dead links as a result.
1 -
Apparently FMP has updated their index with AI and the errors have carried over to FamilySearch:
https://www.findmypast.com/transcript?id=GBC%2F1911%2FRG14%2F01168%2F0585%2F5&tab=this
There are many like this and also found in 1901 Census. I haven't had time to test the others.
2 -
Wow! I've just tagged Find My Past on social media asking if they are aware of the issues. Thanks, @SerraNola
2 -
Thanks for the heads-up @SerraNola !!!
(Insert icon of Jean-Luc Picard doing face-palm…)
I have hit the contact link (not the correction facility) to also report the issue to help emphasise the gravity of the situation.
For those without access to the FMP image, the 5th entry in this household has become the car-crash shown above, because free format comments (which are genuinely on the form immediately above George's name - which is George Walter Cullum - and after it) have been interpreted as part of the entry's name.
From what little I understand about AI pattern recognition, if it comes across a pattern it's never seen before, it just interprets it according to the patterns that it's seen before. If it's a new scenario, then it's up the creek without the proverbial paddle. I have no idea how such "edge conditions" should be treated.
While I've never seen the amount of free format in "Married daughter of…" before, I have seen plenty of "Son / Daughter of the above" insertions.
1 -
The 1911 Census will by far have the most errors because it was produced individually by the householders themselves rather than census takers. Thus, there are numerous annotations in the name fields, and a plethora of handwriting styles. The example I gave was one of the most egregious transcriptions on FMP. Most are more along the lines of these:
Annie Jones Husband at Sea
Just born not married Braggins
Jenny Bates Twice Married
Lodger At Home April Second Blyth
Husband Away From Home
Mr. Walson Mrs Not At Home
The good news is that, for the most part the correct surname shows up somewhere in the surname field so it would likely be retrievable with the FamilySearch search engine:
This one was actually two people, a professional singer named Leslie Dale, living with a married woman (not his wife) named Lydia George. Lydia George would not be searchable.
Children without a surname would likely be found by searching for the parents' surname:
The bad news is that there is a multitude of names with spelling errors that could be irretrievable with the search engine algorithms:
In addition, many wrote the name on the form with surname first followed by a comma but AI did not recognize the comma:
This one was actually "Mabel Elizabeth Roberts" so it would be extra hard to find.
Then there were some given a first name or incorrect name from the row above as a surname:
….or the surname could come from a word in the relationship column in the rows above:
I also found that many of the errors in FamilySearch (that could only have come from AI) had been corrected in Find My Past. This was one that was either user corrected or FMP has already started to fix the problem:
In light of the heavy impact on research, I will find out if anything can be done on our end, but I wanted to make the community aware so adjustments can be made in how search filters are chosen.
3 -
@SerraNola - thanks hugely for those examples. They must be potential problems, but I confess to also finding them interesting!
2 -
Well, I'm afraid that was a waste of my time contacting FindMyPast. I got the response
"... Find My Past has to stick with any original material on the form so we would be unable to change this..."
How on earth those words in the index reflect the name of the person in the household, I have no idea.
Interesting that @SerraNola says that at least one of her car crash indexes has been corrected in FMP. Since all corrections pass through a human (I hope!) at least one of their Support staff has recognised that the index of a name should contain the name and not commentary, in contrast to the response that I got. Sigh.
3 -
I heard nothing back on my social media post. I'll try again.
3
