Baldy wrote:Well,looks like I opened a can of worms with this anomaly ?
The seems to more than myself who is or has had problems with Ancestry
One wonders if anyone from Ancestry reads these thread ?????
![[soapbox]](./images/smilies/soapbox.gif "SoapBox")
![[rant]](./images/smilies/rant.gif "Rant")
, maybe even a severe
warning.....
Wayuullll, you have been warned .......................
One does indeed wonder regarding Ancestry reading these threads, but then I'm picking up suggestions from more than one source to the effect that they're not particularly bothered.
This, in the sense that they are more interested in seeing as many as possible resources made available to their present and possible future subscribers as opposed to paying their sub-contractors a bit more in order to achieve a higher level of accuracy.
There's arguments on both sides
And these have to be put in the context that once you get to an accuracy of, say, 90%, the costs for increasing that accuracy are not linear, but exponential. In others words, and my figures are purely for the sake of example, it could be the case that 99.5% guaranteed accuracy could cost 5 or even 10 times the cost of 90% accuracy.
This also in the context of just what is "accuracy". In terms of the material we're talking about here, it's what was written in the BMDs or censuses.
Especially in relation to censuses, it's well known that the enumerators made errors, some more than others, when they transcribed the info from the individual household schedules to the enumeration books that we know.
I can quote many examples where I'd transcribe the material exactly as it is shown on Ancestry or ScotlandsPeople, but I can prove that it's "wrong", except it's not in terms of the accuracy of the indexing, if you see what I mean. Such an error may relate to the "obvious" interpretation of the handwriting, or parallel, quite unrelated info from other censuses or BMD records.
As I've written previously, I have a certain level of understanding of Ancestry's approach, particularly in terms of, for the Scottish censuses, making more fields available, and allowing more flexible searches in terms of no absolute requirment for the surname field to be completed, and the ability to search on the basis of there being another name linked (on the page , or the household?, - I'm not sure......)
That's not to say that it's not the case that it still regularly happens that I have to be picked up off the floor, and allowed to calm down from my fit of hilarity, and have strong liquor adminstered to me (preferably a 30+ year-old non-chill filtered, single cask, at cask strength, Linkwood
) after seeing the latest example of the complete and utter inanities that are sometimes the output of the Ancestry sub-contractors.
But, as I will never hesitate to admit, "drink taken" is an obvious, occasional problem in terms of the output of census enumerators; as if the fading of ink, - they had to buy their own, and some obviously over-dilluted the ink powder; and that's without handwriting styles, however sober, that would challenge the top experts in terms of interpretation; so that who am I to criticise any indexer/transcriber, either in the UK or elswhere, with or without the advantage of English (or should that be Scots?) as their native language?; and with or without some basic knowledge of Scotland, placename, given and surnames, etc., etc.
That written, my opinions are well known in terms of Ancestry index errors in relation to Scottish county names, including the 18th and 19th century variants, - which all fit on one page; and, beyond that, the 900+ registration district names, - OK, maybe ca. 10 pages of a lookup list .................
Some of these Ancestry county name errors, quite simply, beggar belief; but then, they're balanced by very many others, - the percentage I know not, - that are correct.
But place yourself in the situation of someone working for an indexing sub-contractor, and it's approaching the end of a challenging 8 hour shift, then I'd argue that, despite the quality of the training that you have received, and despite the resources such as lookup lists available to you, the potential for error is considerable.
Did I hear someone say "but what about double entry?".
Aye weel, where cost is not a concern, and this applies to many FHS indexes and the various FreeCensus and similar projects, where a true double entry system is used; it
will result in a substantially higher degree of accuracy.
A true double entry process involves two completely different indexers inputting the data from the same source, with a
third person then checking every situation where the result of those first two inputs is different. It's a bit like good customer service telephone services, - where, if the first person can't resolve the issue, it's automatically referred to a higher level, and so on.
In this case if the third person can't resolve the conflict between the first two, quite independent inputs, then the situation goes up a level to more expert folk, i.e. a
fourth person, and, beyond that, for remaining problematic entries, maybe even another level, i.e, a
fifth person, even more expert as regards the source material .............
But, as can readily be appreciated, that all takes time and costs money.
When you come across a situation where it is claimed that a double-entry procedure is used, then it's worth querying the situation, as you will find, - no names, no pack drills
, - that there are companies who use a somewhat different definition, - this being that one person inputs the data, and a second person then checks that first person's input, and represents that process as being a true double entry process (which it's
not), - a subtle but critical difference from the true definition of the genuine double entry procedure as given above.
The latter definition of double entry might well pick up errors based on what was keyed in, i.e. the interpretation was the same, but the data keyed in was different !!, - but just think again of that 8 hour, or longer shift, and the potential for error deriving from fatigue.
Did someone out there say "But what about quality control?".
I can't speak for Ancestry.
But I can speak for ScotlandsPeople on the basis of a detailed understanding of their process involving sub-contractors, via my membership of the ScotlandsPeople User Group.
This process involves, in the beginning, a 100% quality control (QC) check by GROS of the indexing of a sample of the census involved. Based on the outcome of that initial QC check, further QC is focused on the areas highlighted by the initial QC check.
Having written all that, I'm quite amazed that I haven't even used once the dreaded term "OCR and allied software"
![[5 cups]](./images/smilies/sm-5cups.gif "[5 cups]")
David
