Using deep learning for handwritten text recognition

This Handwritten Gulf Archives project focuses on making handwritten archives about the Gulf region accessible to wider historical inquiry. It begins with transcribing digitized versions of nineteenth-century India Office Records letter books and has been building “ground truth” for the training of deep learning (AI) based handwritten text recognition (HTR) systems such as Transkribus. Its aim is to render such archival collections searchable and usable for study such as in the Historical Text as Data project. This project will publish its HTR model for a diversity of 19th century English hands as well as the text of its letters. It has used theFromThePage platfrom.