+ English Cycling diaries recognised by University of Warwick

We’ve got some terrific results to report relating to an interesting collection of documents held at the Modern Records Centre at the University of Warwick.

Archivist Elizabeth Wood and her team have recently trained a Handwritten Text Recognition (HTR) model to recognise the writing in a collection of cycling diaries written in English during the early twentieth century by David Allan Hamilton.

The pages from Hamilton’s diary are small and frequently broken up with photos, maps and sketches of life on the road.  This meant that the team at Warwick decided to submit a greater number of transcribed pages to train their model.

The Hamilton model was trained on around 200 transcribed pages (containing nearly 20,000 words) from one volume of Hamilton’s diaries.

The automated transcripts produced by this model have a very impressive Character Error Rate of just 5% – meaning that an average of 95% of  characters are transcribed correctly by the computer.

Screenshot of automated transcription in Transkribus. Page from diary of David Allan Hamilton, 1916-1923, from the National Cycle Archive, Modern Records Centre, University of Warwick [document reference: MSS.328/N93/1].
The team at the Modern Records Centre are currently working with the automated transcripts and are also exploring the possibility of training new models to process other diaries in their holdings.

Start unlocking the past with Transkribus

Leverage the power of Transkribus to get the most out of your historical documents.