RootsChat.Com

General => The Common Room => Topic started by: LeedsHipPriest69 on Friday 07 March 25 20:10 GMT (UK)

Title: Poor transcriptions
Post by: LeedsHipPriest69 on Friday 07 March 25 20:10 GMT (UK)
Just been looking on both Ancestry and Find My Past in the hope of finding some evidence of my 4 x Great Grandparents parentage.

I knew that Lancelot Key had died before the first census, but his widow survived him and died a few years later in 1843. Through a copy death certificate I know that Sarah lived in Broad Street Stoke on Trent in 1843, I subsequently found her on the 1841 Census on FindMyPast albeit with the mistranscription of Sarah Kay, easily done.

Anyhow as my main tree data is saved on Ancestry I looked up said person on the 1841 Census, and nothing, a complete blank, so I decided to scroll through the images and as look would have it, no idea how I missed it before, their daughter Elizabeth along with her husband Thomas Tew and family lived in the same area and were even on the same page of the census. The reason I couldn't find Sarah when searching, was that the transcription had her down as Sarah Kong.

You would think with the money both sites extort from us, they would look to review transcriptions, technology having moved on so so much since I first registered with Ancestry.

Gripe over, but I wonder, are there better alternatives out there to Ancestry ? Is My Heritage worth a look ?

Title: Re: Poor transcriptions
Post by: Andrew Tarr on Friday 07 March 25 23:08 GMT (UK)
You have to remember that a good proportion of the transcription was done in India (and maybe other countries) by people unfamiliar with Victorian script.  I have done a small amount myself and recognise the difficulty.  Tho I think a certain amount of checking is done, the cost and delay in checking are significant deterrents.  All end-users like us can do is report errors as we find them, and hope they are acted upon.
Title: Re: Poor transcriptions
Post by: Jon_ni on Saturday 08 March 25 03:29 GMT (UK)
Generally a search of Findmypast will be successful if Ancestry has mistranscribed or there is a missed page and vice versa as most were done independently.
The 1841 can be problematical as was filled in in pencil and has little contrast on the microfilm, whereas all subsequent were in ink.
Certain collections can be better transcribed on one site or the other eg the 1939 on Findmypast tends to have streets carried from the previous page when the top row is redacted whereas Ancestry seems to have instructed their transcribers to just leave blank and not spend the extra time.
The 1841-1901 Scottish census seems better on FindMyPast for the areas I have searched, but one individual/team could be more proficient/dedicated than another. The 1921 is rubbish on both as Ancestry purchased Findmypasts transcripts and they collaborated and jointly did the Irish R.C. baptisms.
Bear in mind you can always search Findmypast even without a subscription and the results summary with parish and district for census may be sufficient to then locate the mistranscription on Ancestry eg by eliminating surname but including birth year and birth county. They generally show more than Ancestry does without a sub in their search index results so is where I turn to.
In your Sarah Kay/Key example you have knowledge from subsequent or prior records to assist and can spent time zooming in and comparing how an individual enumerator formed his letters, rather than reading in isolation with targets to meet and hence time pressure to get the page and ledger completed efficiently.

A.I. is now being used by Ancestry and Familysearch for bulk transcripts of handwriting eg 1950 US and Familysearch have implemented on some previously untranscribed films https://www.johngrenham.com/browse/retrieve_text.php?text_contentid=524 certainly faster than the graph on https://irishdeedsindex.net/
Title: Re: Poor transcriptions
Post by: HughC on Saturday 08 March 25 06:53 GMT (UK)
It annoys me that when I submit a correction to the 1921 census to Ancestry, it is shown only as an "alternate": a search using the correct spelling is still just as unlikely to find the person.  So why do I bother making corrections?

In my experience you can forget My Heritage.
Title: Re: Poor transcriptions
Post by: wilcoxon on Saturday 08 March 25 07:12 GMT (UK)
It annoys me that when I submit a correction to the 1921 census to Ancestry, it is shown only as an "alternate": a search using the correct spelling is still just as unlikely to find the person.  So why do I bother making corrections?

In my experience you can forget My Heritage.

That is so true, I have been looking at a lady with an unusual middle name and in all cases she is only showing as "alternative".  As her other names are Mary and Jones there's little chance of others without knowledge of the siblings finding her.
Title: Re: Poor transcriptions
Post by: susieroe on Saturday 08 March 25 07:53 GMT (UK)
When we were transcribing the 1851 Census we were told to enter what we could see, not what we thought it should be; ie., between 'tailor' and 'sailor', if the t definitely looked more like an s then you'd plump for sailor, even though it was in deepest Leicestershire and you would assume it was tailor. 3 people transcribed the pages, which a fourth person then scrutinised and approved the best copy.
Title: Re: Poor transcriptions
Post by: HughC on Saturday 08 March 25 08:10 GMT (UK)
An important reason for transcribing is to create an index so that one can find people.
Therefore it makes sense to transcribe what was obviously meant, not (for example) what an enumerator mistakenly assumed was written.

And transcribers should have been chosen who had at least some idea of names of people and places.
Given that that was not done, corrections submitted by relatives should be accepted.
Ancestry's handling of the 1921 census is a disgrace.
Title: Re: Poor transcriptions
Post by: MollyC on Saturday 08 March 25 09:42 GMT (UK)
Are Ancestry's suggestions indexed?  I doubt it.

Findmypast generally deals with corrections to the 1921 census, and other record sets, though if they reject a correction it would be helpful also to have space for suggestions.  It would be also be useful because I found they cannot deal with corrections to record sets transcribed elsewhere and purchased by FindMyPast, eg Canada censuses, so there is no button under More Actions.  FindMyPast said they would pass my corrections on, but they have not been acted on.

What makes no sense in Ancestry is finding members of one household with different surname spellings, because a transcriber wrote what they could see and it was not scrutinised.  FindMyPast's competance in indexing is generally more satisfactory.
Title: Re: Poor transcriptions
Post by: Andrew Tarr on Saturday 08 March 25 12:00 GMT (UK)
An important reason for transcribing is to create an index so that one can find people.  Therefore it makes sense to transcribe what was obviously meant, not (for example) what an enumerator mistakenly assumed was written.
Having done a fair bit of transcribing, I sympathise with your sentiments, but the difficulty is in deciding how 'obvious' the meaning appears to be.  In the above example of choosing between Tailor and Sailor, or Lawyer and Sawyer, one has to 'transcribe what one sees'.  That can be a moot point, especially when the transcriber is unfamiliar with the enumerator's way of writing.

However I can see no excuse for entering a ridiculous version of a well-known place name, for example, just because the original writing seemed unrecognisable.  It should be worth a minute's deliberation.  It is also impossible to argue about the spelling of surnames, which have no truly accepted spelling.  One example : I have transcribed church records for a rural Lancashire parish, where the same vicar alternated the spelling of Howarth and Haworth, apparently at random, for about 50 years.  I very much doubt that the various families insisted on one version or the other.
Title: Re: Poor transcriptions
Post by: Jon_ni on Saturday 08 March 25 12:10 GMT (UK)
Quote
transcribers should have been chosen who had at least some idea of names of people and places

Agree but the reality is cost is the major factor. Findmypasts long-time transcription partner is SBL in India. https://www.facebook.com/share/v/15sTteZZmM/
These days a QC consists of a 10% random check not 3 people and a scrutiniser and 98.5% accuracy.

As the 1921 was done before the 100 years were up, like the 1939, "each digital image was broken up into segments so that the person transcribing it could not see a whole record or household in its entirety." Perhaps explains surname differences for some children - transcribed by different people.
https://www.nationalarchives.gov.uk/about/freedom-of-information/information-requests/1921-census-transcription-accuracy/
Ancestry's search algorithm picks up user submissions and also what others have saved to trees hence suggestions of 1921 from 1911 & 1939 and the rubbish suggestions because someone has erroneously selected another one in a different country yet has others the same as you.

Alternate corrected information (names, location) on Ancestry is incorporated such that it aids others who later search to locate the person and item much more quickly - but is not implemented the instant you add or update the info. The comment
Quote
It annoys me that when I submit a correction to the 1921 census to Ancestry, it is shown only as an "alternate": a search using the correct spelling is still just as unlikely to find the person.
is untrue, it is of benefit to others.
Title: Re: Poor transcriptions
Post by: Ayashi on Saturday 08 March 25 12:17 GMT (UK)
I spent a long time looking for my CARTER family once... it wasn't until I went through the original pages that I found them. They were transcribed as LEOTA! I had a similar situation with MILBURN, eventually finding Thomas MILLIBUN living with his father John MILLIRON  ::)
Title: Re: Poor transcriptions
Post by: Nick_Ips on Saturday 08 March 25 14:28 GMT (UK)
...
Given that that was not done, corrections submitted by relatives should be accepted.
Ancestry's handling of the 1921 census is a disgrace.

as far as I know user submitted 'corrections' are accepted and become searchable, albeit as alternates rather than replacing the original entry.

I don't see this being a particular problem, provided the search algorithm does pick up on user generated alternatives.  If we treat the transcriptions as a finding aid (and having spent days and days at the FRC winding forward and back through microfims I'm delighted to have whatever help we can get) then it isn't critical what the transcription says, so long as the right record can be found.

I have to say, rather than finding it a 'disgrace', I think Ancestry's approach to the 1921 census has been pretty good. Yes, it would be nice if they had retranscribed the whole lot, but just having an alternative means of searching the same data has been valuable to me.

I've found many 1921 households since the Ancestry launch that I'd spent hours looking for on FindMyPast with no joy - in particular by searching Ancestry for a different recordset for an individual (e.g. 1901 or 1911 census) and then seeing what gets listed in the 'suggested' records for the 1921 census.

Ancestry's approach to fuzzy searching - in my view - is delivering better results than FindMyPast.  If that is a byproduct of Ancestry having an awareness of inaccuracies in their data and developing a search algorithm to compensate then I'd say that was more than fine with me.  :)  Getting a result from a finding aid is as important (more so IMV) than every record being transcribed with 100% accuracy... because, let's face it, the original data captured in 1921 was far from perfect.
Title: Re: Poor transcriptions
Post by: GrahamSimons on Saturday 08 March 25 14:39 GMT (UK)
I've been working on the SoG Great Card Index. This is unlike a census or a parish register as the images come from a wide variety of sources and don't therefore provide hints. There are BMDs and MIs but also everything from pedigrees in various repositories,  sedan chair licences  in Dublin, books about the East India Army, Yorkshire clergy, RN promotions, MS collections,  Close Rolls, PCC will transcriptions...... Many of these have already been transcribed once to get into the Card Index, some more than once.  And some of the handwriting is spectacularly illegible! Add to that variable spelling over the last 800 years or so, and random choice of spelling of Welsh placenames by English speakers..... We are doing the best we can and there's quite a collegiate approach to problem solving. 
If you'd like to experience this, there are volunteer opportunities (and you don't need to be a member, although membership brings lots of benefits. )
Title: Re: Poor transcriptions
Post by: wilcoxon on Saturday 08 March 25 18:01 GMT (UK)
Quite a few years ago I helped transcribe County parish registers for a local history societies publications.
This was in Wales.  Many placenames were so bad, but I knew that what looked like Riddlin was Ruthin, Rubban was Ruabon etc because they were parishes, and that's before we even started on the smaller villages.

We still had to write what we saw, not what we  knew it  to be, this was all cross checked to make sure that the  transcripts were accurate, even though actual spelling was wrong.
I was told that the searcher was supposed to decide for themselves if it was correct.
Thankfully in the published paper booklets alternative name spellings were filed together with similar names, for example Jones,  Jonas, Johns, etc.
Still not much use when searching online.
Title: Re: Poor transcriptions
Post by: coombs on Sunday 09 March 25 13:21 GMT (UK)
When I started becoming a transcriber I was told to transcribe what you see. I know that may sound a cliche but only going by what the bosses want.
Title: Re: Poor transcriptions
Post by: Andrew Tarr on Sunday 09 March 25 23:05 GMT (UK)
When I started becoming a transcriber I was told to transcribe what you see. I know that may sound a cliche but only going by what the bosses want.
This is clearly the ideal rule in a perfect world, but when the script is apparently 'meaningless', should the transcriber not be allowed some latitude in making a rational interpretation ?  I always take the view that rules are allowed exceptions just to prove them from time to time.  The purpose of transcribing is to make records available to enquirers, not simply to create an exact copy of whatever characters can be made out in a puzzle document.
Title: Re: Poor transcriptions
Post by: HughC on Monday 10 March 25 06:51 GMT (UK)
The purpose of transcribing is to make records available to enquirers, not simply to create an exact copy of whatever characters can be made out in a puzzle document.

Thank you, Andrew.  Unfortunately Ancestry (and to some extent FindMyPast) haven't grasped that simple truth.
Title: Re: Poor transcriptions
Post by: MollyC on Monday 10 March 25 09:06 GMT (UK)
When we transcribed the 1881 census, after it was fully released, it was organised by local FHS groups so there was local knowledge.  It was transcribed twice by different people and was "transcribe what you see" but we were encouraged to make notes on the back of the sheet.

I had one family with children born in:
Carcroft - straightforward, about 30 miles from the residence place
Skyler - not in any gazetteer - but the next village to Carcroft is Skellow
Title: Re: Poor transcriptions
Post by: GrahamSimons on Monday 10 March 25 09:18 GMT (UK)
Returning to the SoG Great Card Index, we have great difficulty when we can 'transcribe what you see' but it makes little sense. Often we are getting placenames from the 14th century which simply can't be found now - written in neat hadwriting (for once) but without a county or other details to give a  clue. As for personal names with dreadful handwriting all we can do is make a best guess - not so much 'write what you think you might see if only this person could write even slightly legibly.'
I'm reminded of the rubric on my university exams which said, "Candidates are advised to write legibly. Those who do not may place themselves at a severe disadvantage."
Title: Re: Poor transcriptions
Post by: Andrew Tarr on Monday 10 March 25 09:29 GMT (UK)
I had one family with children born in:
Carcroft - straightforward, about 30 miles from the residence place
Skyler - not in any gazetteer - but the next village to Carcroft is Skellow
If the writing is legible there is no problem with following the instruction, clearly.  When it's not, then transcribers are on their own.  Your example shows that early enumerators were in a similar position - they wrote what they heard rather than what they saw.  That's why surnames have so many variants; by the time many people could write, those variants were established and their owners insisted on keeping them.
Title: Re: Poor transcriptions
Post by: Sloe Gin on Sunday 16 March 25 19:42 GMT (UK)
I think Ancestry's approach is by far the most practical and useful.

The main aim is to get to the original image, where we can use our own judgement. The index is principally a finding aid, so recording every reading or version of an entry in the index improves our chances of finding someone whose name may have been misspelt or just wrong.  When you search for someone on Ancestry, the results will include any alternatives that have been contributed.

Only yesterday I benefited from this while looking for a Romani person in the censuses. Romani can be particularly hard to find as some have unusual forenames which can be recorded in a variety of spellings, and some went by more than one surname, using that of either parent at different times, and men sometimes used their wife's surname.

Findmy past could only find my man on one census, but thanks to the alternatives Ancestry gave me three.
.

Title: Re: Poor transcriptions
Post by: Sloe Gin on Sunday 16 March 25 19:48 GMT (UK)
Are Ancestry's suggestions indexed?  I doubt it.

Yes, they are. They wouldn't be of any use if they weren't.
Title: Re: Poor transcriptions
Post by: MollyC on Sunday 16 March 25 20:20 GMT (UK)
Unfortunately Ancestry does not make that clear.  I have hesitated to make suggestions because it seems they are only visible after you have arrived at the answer anyway.  Where a suggestion is definitely the better solution they could help by changing the main transcription and leave the original interpretation under suggestions as a "previous transcription" or similar.

As I said it would also be helpful if Findmypast had space for (indexed) suggestions.  I am stuck with the last line of a Canada census page where the surname has been copied from the family above, instead of being transcribed as a different single person household.  The RH side of the page shows it is a different building.
Title: Re: Poor transcriptions
Post by: Nick_Ips on Sunday 16 March 25 21:11 GMT (UK)
...Where a suggestion is definitely the better solution they could help by changing the main transcription and leave the original interpretation under suggestions as a "previous transcription" or similar....

But that would make no difference to finding a record, provided both (all) versions are indexed.

Deciding which record was definitely the better solution would take resources (=cost).  I'd rather Ancestry use the resources they have to give us access to more record sets, or else keep the subscription costs down.

Achieving greater accuracy in the transcriptions of records that have already been described doesn't really gain us much, so long as there are tools to find the records in the first place.
Title: Re: Poor transcriptions
Post by: Sloe Gin on Sunday 16 March 25 21:23 GMT (UK)
Unfortunately Ancestry does not make that clear.  I have hesitated to make suggestions because it seems they are only visible after you have arrived at the answer anyway.  Where a suggestion is definitely the better solution they could help by changing the main transcription and leave the original interpretation under suggestions as a "previous transcription" or similar.

There's no need for any of that.  Any of the suggested transcriptions will bring up the same result as the original transcription. That's all you need. 

To give my example, the person I was looking for was named Lock, but on some records he is Boswell. Because someone had kindly added Lock as an alternative, the Boswell entries came up in the results as well as the Lock entries.

Quote
As I said it would also be helpful if Findmypast had space for (indexed) suggestions.  I am stuck with the last line of a Canada census page where the surname has been copied from the family above, instead of being transcribed as a different single person household.  The RH side of the page shows it is a different building.

Yes, there are lots of cases like that. I have an entire family entered under the wrong surname in one census.  I've added the correct surname to Ancestry so they can be found now, but didn't bother with FindMyPast as I knew they would not accept it.
Title: Re: Poor transcriptions
Post by: Ayashi on Wednesday 19 March 25 22:49 GMT (UK)
Just spent a while looking for a DAVISON on census. Eventually found under DANISCH. Silly me  ::)
Title: Re: Poor transcriptions
Post by: Sloe Gin on Wednesday 19 March 25 22:59 GMT (UK)
Just spent a while looking for a DAVISON on census. Eventually found under DANISCH. Silly me  ::)

If this is on Ancestry, hope you've filled in the correction.
I had DAIRS for DAVIS once.  It's still there, but DAVIS finds it now.
Title: Re: Poor transcriptions
Post by: Ayashi on Wednesday 19 March 25 23:00 GMT (UK)
It was on Anc, and three other people already had  ::) I had to find it on FindMyPast first (trans correctly) then use various details to eventually find it on Anc.
Title: Re: Poor transcriptions
Post by: Sloe Gin on Wednesday 19 March 25 23:10 GMT (UK)
That's strange. Once a correction has been submitted, it usually bypasses the wrong transcription.