About the project
The Rothschild Archive holds 325 volumes of handwritten Cash Books as part of the London firm’s Cashiers department. Next to private expenses such as visits to the hairdresser or donations to charity, the books record government bonds and further corporate investments. In cooperation with the University of Regensburg, the Cash Book project aims to index and visualise latent information in the books using new technology within the field of Digital Humanities. Artificial Intelligence and Machine Learning can be employed to recognise, annotate and analyse historical handwritten text.
Handwritten Text Recognition using Machine Learning is one of the first building blocks of the project. Unlike classical programming which requires specified rules to perform functions, machine learning needs to be taught with what is called Ground Truth. To create a Ground Truth data set for Handwritten Text Recognition a portion of the Cash Books needs to be transcribed manually. By processing the transcriptions and the corresponding digitised images of the Cash Books, artificially intelligent systems can learn to recognise the handwriting and convert it into computer-readable text for research.