Our Transkribus platform offers access to cutting-edge Handwritten Text Recognition technology capable of processing all kinds of historical documents. Users upload digitised images of historical material to Transkribus and transcribe each page to create training data for Handwritten Text Recognition. A training set of at least 15,000 words or around 75 pages is enough for our algorithims of machine learning to recognise the writing and produce transcripts suitable for research and full-text search.
The process of training a Handwritten Text Recognition model takes place on our servers – but Transkribus users can now set up and monitor the training process themselves for the first time. The power is in your hands!
You can find out more information about how to train your own model in our new How to Guide:
The training function will be included as a standard feature in future versions of Transkribus. If you would like access to the feature now, simply send an email to the Transkribus team.
The great thing about Handwritten Text Recognition technology is that it learns from all that it sees. As more and more images and transcripts are submitted to our platform, the accuracy of the recognition will only increase. So we are looking forward to lots of new models and experiments!