Entering Ancestors on Lost Cousins website

Well done, Anne - let's hope that your cousins follow your excellent example, so that before too long you get the matches that you deserve.

 

Hi Charmaine,

FMP doesn't appear to have the Book reference listed. However I have found the Piece/Folio/Page on Ancestry but I need the names to determine which book you need.

PM me if you don't want to put the names in public.

 

Charmaine, this FAQ on the LostCousins site explains how the 1841 census references work.

 

Peter, the book numbers for piece 769, folio 4, page 3 don't appear on the scan at FMP. However Ancestry do have something recorded.

 

I think it depends on the area. In some areas FMP include the book number in the folio number eg 8/12 means book 8 folio 12.
From 1841 census at FMP
On transcription page:
Piece: 484
Book/Folio: 8/6
Page: 5
Above image: HO107 piece 484 folio 8/6 page 5

 

The book number is missing from approximately one-third of the counties in findmypast's 1841 England & Wales Census, ie there are whole counties with book numbers and whole counties without. In almost every case you can find the missing book number on the image, which you should be using for this census (at LostCousins we only use transcribed data in 1880/1881).

Where the book number is missing from the image (and can't be found from pages before or after in the same book) Ancestry is the best source. You don't need a subscription - as long as you have the other census references you can deduce the book number using a "trial and error" approach.

Note that the Ancestry census reference data is unreliable - the folio number is often missing, and where given is often wrong. Always check the references against the image using the guide I linked to above.

 

I always use my own transcription where I have access to the original image. I don't understand why you need the transcription for the 1881 census since the census reference is the same whatever the transcription is. I do put in the 1881 transcription if I still have it in my file but for many people I've replaced it with my own.

 

The reason we use the transcription in 1881 is so that everyone enters the same data whether or not they have a subscription to Ancestry or findmypast.

If what you enter differs from the transcription then it won't match with your cousins' entries.

If the difference is very slight then you might find that the entries are flagged with a red ! to indicate a near match, but that can't be guaranteed. However if this does happen, go to Ancestry and make sure that what you've entered matches the original transcription (you don't need a subscription to do that).

Always follow the advice on the Add Ancestor form to ensure that you have the maximum chance of finding cousins (and remember that it changes depending on the census you're using).

 

Then you're not matching on census reference.
On this argument we're unlikely to find matches in other censuses as we are using different transcriptions. You appear to be saying that the match code you use will only match us if we've hit on the same transcription. This is disappointing as I really like the concept of matching censuses. It was a very clever idea to match on census reference but if you are really matching on name why bother with the census reference? I surely must be misunderstanding how it works.
Surely a matching census reference should show some sort of match even if the name doesn't match. People may have entered someone in the same household even but a different name. For example they may not have included servants or boarders OR they may only have entered the servant or boarder of interest OR only entered their child of interest and your child of interest is a sibling. Yes I know we're encouraged to add the whole household and I mostly do but I can't guarantee others will.
I certainly won't be wasting research time in order to find inaccurate transcriptions.
Very discouraged from making any more entries.

 

I suspect that the algorithm does match on census reference but then requires the names to match. If you think about it on any one page there will be up to 22 entries often 7-8 households. You wouldn't necessarily want a match with someone first on the page if your entry was last on the page in a different household. Although knowing you matched someone on the page might be helpful if the other person then knew something more about the area.

I agree that requiring a name match does weaken the chances of a match and makes it less likely a match will be made if people aren't religiously entering exactly what was transcribed even if it is wrong. I suspect that a whole lot more matches would be made if the requirement to match on a transcribed name wasn't included.

This could be achieved in a number of possible ways :

1. The easiest option would be to show "page matches" ie: you match a census reference but not on names. This would need to be highlighted as not a 100% match unlike the nature of all the other matches. This may be what the "!" option does although I suspect it is checking for similar names rather than just that you match a page with someone else. If this option was available then ideally the match (for paid up members only) would show the surnames entered on both sides, so you could visually check if it was just a mistaken entry or a completely different name. NB. Humans are dramatically better at making these sorts of judgement calls than computers.
2. Have an optional box to record the household number on the page. 1 = 1st household, 2 = 2nd household etc. Note start at 1 on the page regardless if its part of the household on the previous page. This would allow additional matching so in your example of two children only one entered on each side they would match as they would share the same household number. This would give similar 100% matches to the existing system. Note the box would need to be optional as a) it hasn't been filled for the existing records and; b) because not everyone will see the images and know what household number on the page it was. This would mean you had matches purely on an "extended" census reference thus the same sort of 100% match as the existing system but with the added benefit of not relying on the name matching.
3. Taking idea 2 further and having an optional box to record the line number on a page. This then extends the census reference further to the exact line on a page. Again this could be promoted as an "extended" search that offers greater matching and not a mandatory field. Note that this sort of match only works for people who enter everyone in the household they are interested in.

Thus option 2 is probably the simplest to extend the matching and enhance the matches where the name matching is proving dodgy. However that requires extra work on the data entry persons behalf and not everyone will have access to the census page image. Option 1 would be easiest to implement as it doesn't require the end user to enter extra data. It would however add a category of "page match" rather than the 100% match we have at present.
Technically neither of these are in any way difficult although I understand that the person who originally wrote the code is no longer around so changing things like the matching algorithm may prove tricky for Peter to implement.

 

The primary match is on census references - if they differ then you won't even get a near match. But we can't just match on the census references because in most censuses they define a page, not a household, and certainly not an individual. With up to 25 people per page of the 1881 Census and an average of 4.3 people per household about 80% of the matches would be wrong if we matched simply on the references.

Both Ancestry and findmypast licensed FamilySearch's transcription of the 1881 England & Wales census, rather than producing their own - that's why it was feasible as well as desirable to use the transcription rather than the handwritten schedule.

Matching is by individual, not household. It doesn't matter whether both members have entered the same people as long as there's at least one individual they've both entered. Because most people enter every relative in the household it's quite unlikely that two cousins would both have entered only one person, and even where this is the case, they'll usually have entered the head of household - in which case they'll still get a match.

The same system has been in use since LostCousins started over 9 years ago, and the instructions on the Add Ancestor form haven't changed (although the wording has probably been revised a little over the years).

IMPORTANT: enter the following information exactly as it appears in the census transcription even if you know that it is wrong or incomplete. Do NOT take information from the handwritten census schedule, as it may prevent your relative being matched with the same person entered by another member.

If it is any consolation, for you to check all of your 1881 entries and amend them where necessary would take significantly less time than it has taken me to write this reply. But I don't consider it a waste of my time, so hopefully you won't consider it a waste of yours either.

 

Unfortunately no matter how easy it might be to make the programming changes (and it wouldn't be easy, because it would change the fundamental basis of the matching system and the database structure), persuading the tens of thousands of members who have already made millions of entries to go back and add extra information would be impossible!

Back in 2003 I did consider options 2 and 3. In all I spent 6 months working out how LostCousins might work before programming started - longer than it took to write the software and launch the site (although obviously development continued after the initial launch as a result of feedback from users and new ideas that came to mind once the site was in operation).

Those options were not feasible because they required sight of the handwritten census schedule. In those days only Ancestry had the census images, whereas the transcription was and is available free online. However, even if that hadn't been a problem there's too much scope for miscounting - it's not always clear where one household ends and the next begins, there can be empty properties, and there are often blank lines on the schedule.

Option 1 was something that I briefly considered after launch as an alternative to introducing the Neighbours feature. However, the very small increase in the number of genuine matches would have been miniscule compared to the vast number of incorrect matches that would be generated. Since LostCousins is the only site to offer 100% accurate automated matching - it was my frustration with Genes Reunited's inaccurate matching that prompted me to look for a better solution - there was no point changing.

 

Yes, it could be a mistyped name, but it could also be an initial or middle name that has been inadvertently added (or omitted). Occasionally it could be a red herring, for example if there are twins called Jane and James. Also, see below....

Findmypast and Ancestry both licensed their 1881 transcription from FamilySearch (that's why it's free - FamilySearch won't let them charge for it). A few discrepancies have subsequently developed because of the way that findmypast correct transcription errors, but there are so few that the chances of a match being completely missed as a result really are miniscule.