Further Work on 17th Century Shorthand – Machine Learning for Palaeography
David Rogers explains our continued historical shorthand research work.
Unfortunately I’m not in the US this year for the hacking conferences but I hope that next year I will be able to go and explain and demonstrate more of our groundbreaking work on reading 17th century shorthand and other logographic languages.
As part of our Copper Horse shorthand AI model building work, we’ve been acquiring and saving a lot of shorthand manuals and material that can help us. Many of these were destined for destruction. The acquisitions have led to us having one of the largest shorthand collections in the UK with some really important works in there.
This evening I received one of the very few remaining copies of William Addy’s 1687 shorthand bible. This was almost certainly going to end up being destroyed as it is in a poor condition. We are arranging for its repair and will be able to display this for people to see afterwards. A small sample of Addy’s writings below (his was a type of ‘Semigraphy’).
These early shorthands were very different but looked very similar because of some re-use of letters and just the limit of how many line strokes you can make. They are extremely hard to learn and understand. One of the biggest challenges for us is the “arbitraries” – essentially reserved words, often which have no proper reading logic to them – they needed to be memorised by the writers. Some of them did have better logic but were still a memory test – for example the row of dots in the image above means ‘innumerable’. Many characters also clashed with other words or phrases with the same symbols (it was still early days in the development of shorthand in general!). In Addy’s system, you find a lot of arbitrary complete phrases – this meant that you could write even longer documents in very small spaces, but that it was also incredibly difficult to decipher! In the manual for the system you end up with a special symbol for such common phrases as “to run headlong to hell” which is conveniently noted as also the same for “to fall headlong into the pit”! The character that represents this looks like a picture of a candle. Thankfully shorthand sorted itself out a bit in later centuries, but we still have many documents written in a multitude of 17th century systems that are still not read. There are a tiny handful of people who study this domain and who can read any of the shorthands at all. I’m hoping that we can provide tools that can assist in that.
At the bottom of the text below William Addy has signed his name with the strokes of his alphabet and the vowels represented by positions around the consonant. This is all phonetic rather than directly alphabetic, so you have to say some of the words out loud to understand them:
W(i)L(i)M ADi
This breaks down as follows:
The ‘L’ shaped symbol : W
The ‘U’ symbol is then in the position of (i) as vowels surround the consonant in various positions. This symbol represents the letter L connected in the (i) position to a line which represents the next letter ‘M’.
For the surname:
The ‘forward slash’ symbol is the letter A, connecting to the semi-circular ‘D’ and then a dot for the ‘i’ on the end – representing the sound of ‘y’.
