The British Library, one of the READ project’s Memorandum of Understanding partners, has been working with Transkribus to process records from the India Office. This collection relates largely to the London-based administration of the East India Company and pre-1947 government of India.
The British Library started to experiment with Transkribus technology back in 2015. The complex layout of some of the documents and the number of different hands means that this collection represents a challenge for automated processing. But the latest results show that an Automated Text Recognition model can transcribe pages with a satisfactory Character Error Rate (CER) of 15%.
Alex Hailey, Curator of Modern Archives and Manuscripts explains more about the progress and lessons learned in his blog post on The British Library’s Digital Scholarship blog.