FreeBMD-Admins-L ArchivesArchiver > FreeBMD-Admins > 2000-06 > 0961551188
From: Dave Mayall <>
Subject: Re: Accuracy in FreeBMD
Date: Wed, 21 Jun 2000 02:33:08 +0100
On Tue, 20 Jun 2000 19:53:20 +0100, you wrote:
>Whilst this submission is not intended to be critical of FreeBMD, I have
>wondered what the rationale behind 'FreeBMD is a *STRICT* transcription
>of the indexes, and it is important that the source data is never
>embellished or corrected'
[snipped commentary on the extent of errors within the index, which I
Whilst it may seem paradoxical, the key to ensuring that FreeBMD is
ultimately *more* accurate than the source we transcribe from lies in
faithfully transcribing the original errors, if we "correct" entries
as we transcribe them we risk introducing more errors (as Mike Foster
notes, the typing of handwritten indexes has already done that).
The point at which we can start making corrections is when we have all
the data for a quarter, such that we can make our corrections in the
light of having all the data in front of us.
It is important that when we do come to work out where the errors lie,
and "improve" the data, we clearly identify what was source, and what
we have interpolated. That way the corrections, if proved wrong, can
be easily identified and undone, because we *know* which bits are
corrections. If some of the "original" data is modified (with the best
of intentions) we will never be able to reconstruct it.
I will offer again and example I have given before (and it *is* a real
example encountered by a member of my syndicate)
Marriage December 1854, BENNETT, James, Hatfield, 7b, 1066?
The source can be seen at;
There is obviously an error here, Hatfield isn't in volume 7b, it is
in volume 3a, yet the original index is quite clearly Hatfield,7b. The
transcriber could quite easily have "corrected" this entry to
Hatfield,3a (it's obvious that it's an error isn't it)
This is an excellent example of a correction which would have been a
big mistake. Closer examination shows an alternate explanation.
Hatfield may be volume 3a, but Hayfield is volume 7b a far more likely
error to have been made by the indexing clerk.
With the benefit of an accurate transcription of the original entry,
and the ability to analyse page ranges, we can ultimately annotate
this entry such that a search will tell us
BENNETT, James, Hayfield, 7b, 1066? [Index has District=Hatfield]
If the transcriber had made the correction, we would simply have had a
rogue entry in volume 3a which probably didn't correspond to the page
numbers of the rest of Hatfield, and which, without the evidence of
*exactly* what was in the original index would have been far more
difficult to resolve.
In summary, the transcriptions need to be literal, but we will use
analytical techniques to present the searcher with an *opinion* where
there is evidence that the original was in error.