Curtis’ Experience of Transkribus:
On the whole, I’ve found Transkribus incredibly accessible and easy to use, with all the necessary tools for this project. I started the project by manually transcribing an assigned extract from the diary. This was simple enough and was a good ‘warmup’ for the rest of the project… Once enough of the diary had been transcribed manually, it was possible to train the handwriting recognition tool. While making the model included a few confusing steps, it worked out fine and a model of Upcott’s handwriting was successfully made. We applied this model to the rest of the diary and managed to get a decent transcription of it all, though proofreading was required.
Tranksribus has been very useful, and I would urge anyone with a similar project involving manuscripts to use it. With help from some of the Transkribus team in Europe to begin, I was able to learn how to use the software very quickly. I developed my understanding of the software throughout the project after being equipped with its foundations. This was largely catalysed by its accessible user interface, which meant that tools and functions were easy to find. Now that I’ve been trained in using Transkribus, I would feel comfortable doing other projects involving the digitalisation of original manuscripts.
With further material to be transcribed, the handwriting model on Transkribus made by the team will continue to be updated. Aside from Upcott, other projects have already created handwriting models for some manuscripts - a model of Jeremy Bentham's handwriting can be found, as well some non-English models too, such as Charles V of Spain (Early Modern Spanish) and Konstantin Rychkov (Russian). There are also models of typed writing or general writing from certain eras (for example, 'NLF_Newseye_SV' is trained to work on Swedish script from the late eighteenth to mid twentieth century).