New Sherborne Swans and Secretarial Script
Cursive necked swans and cursive transliteration.
I’m pleased to report what some of you have noticed - there appears to be a “new” pair of swans on the brook. One is a fully white adult, the other is a cygnet but of noticeably large size. The swans currently transit between Haycroft and the top end of the Broadwater - let me know if you see it on the Broadwater or the Lower Narrow-water. Here’s some pics taken by Rosi.
Nina at Haycroft reports that the adult is quite shy but the youngster is much more comfortable with nearby humans. Sexes are not yet identified.
I’ve had some queries from subscribers about how my transliteration of the 1712 Sherborne accounts is going. As a reminder this document written by John Dutton in 1711 -1712 details the management of the Sherborne estate in the years immediately after he took control of the estate from his father Sir Ralph. Sir Ralph seems to have been banished in disgrace to Ireland. The accounts consist of hundreds of pages like this:
The truth is that it is taking me much more time than expected. I use AI a lot for other analysis (such as the translation of latin charters) but AI is less good at interpreting the cursive script than I had hoped - it guesses and frequently the guesses are wrong. So I’m having to spend more time on it and time, like everything else, is in short supply.
I have found that I needed to learn much more about the script being used. If you are interested, there’s some good online training course, really useful, to addresses what is known as “Secretarial Script” which is a bit older than the accounts, but is broadly similar - just recognising the odd ways letters were written has helped a lot. But still don’t expect results from me on these accounts for a year or so
If the bug catches you I’m more than happy to share the workload, and I’ll share the training courses and reference documents I’m finding useful, below. Once we have the accounts fully transcribed we can apply some analytical tools to it and I’m sure it will give us tremendous insight to the history here in Sherborne three hundred years ago.
But the rapid pace of AI means that there are AI models being “trained” which promise to do a better job and I’m currently giving this a go, but again with limited success. Here’s a simple example:
I uploaded this text
Now I can see myself that this reads:
September 19th 1712
Rec’d of Fowler of Morton
balance on account £6
Rec’d of Jer. Bulcher(?) on
account of Rent £14
But one of the best on-line AI models for script reading from this period called “Egerton” gives me this from the same text:
So that’s pretty rubbish! The advantage of doing this in person too, is that one can apply contextual knowledge such as local place names and surnames, which helps enormously.
