Potentials of Advanced Text Technologies: Machine Learning-based Text Recognition (PATT)

Automated Text Recognition (ATR) is a method that uses algorithms or artificial intelligence to convert handwriting or old printed text into computer-readable text. This enables researchers to analyse and process large quantities of handwritten pages. At the same time, the method helps laypeople and non-specialists to decipher old scripts or documents. For automated handwriting recognition, the relevant programmes require a digital version of the original documents and a model that is as similar as possible to the script to be transcribed. These models, in turn, are trained and customised by researchers and institutions using large amounts of data for various purposes.

Over the course of the two-year project, the ‘Potentials of Advanced Text Technologies: Machine Learning-based Text Recognition (PATT)’ project established a panel of experts and organised regular workshops. These enabled the project team and others working with ATR to exchange the latest research findings and adapt best practices in the use of ATR to rapidly evolving technical possibilities. The second successfully achieved objective was the development of a learning module on the use of ATR. The learning module is freely available to all interested parties on the Ad fontes Open Access e-learning platform. It facilitates the use of ATR in line with current technical capabilities for a wide range of researchers and interested non-specialists.

The expert panel and workshops will be continued on a rotating basis by the institutions that have participated so far. The Ad fontes e-learning platform will be progressively geared more towards the use of ATR in the coming years.