Poor transcriptions

[Jon_ni]

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

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.

[Ayashi]

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 

[Nick_Ips]

...
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.

[GrahamSimons]

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. )

[wilcoxon]

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.

[coombs]

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.

[Andrew Tarr]

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.

[HughC]

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.

[MollyC]

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