A research team led by CMS professor Michael Gervers has developed software to read and transcribe handwritten Latin. The new software will launch virtually on March 3, followed by a globally accessible workshop on March 10.
The software is based on Transkribus, a program for artificially intelligent text editing. Transkribus can be trained to recognize any form of handwritten text and image patterning, enabling rapid transcription and cross-comparison between texts. The program has the capacity to revolutionize text editing, as it not only transcribes but can recognize "hands" and identify cross-textual patterns and similarities.
Gervers's team, including the DEEDS project, has trained Transkribus specifically using 13th-century Latin legal texts from England, though they have also collaborated with the Bentham Project (University College, London), which is working on handwritten 18th-century English. By sharing costs for software development, the two teams have each perfected Transkribus for their individual needs.
Members of the Bentham Project, as well as the DEEDS project, will participate in the launch and workshop on March 3 and 10.
During the initial machine training, researchers fed images and lines of sample text into Transkribus, specifying how each word was to be transcribed. Based on the sample lines, the program extrapolated rules for transcription, including anomalies like hyphens and abbreviations.
Though Transkribus currently is optimized for 13th-century Latin legal handwriting, Gervers notes that this is only the tip of the iceberg. "With more investment of time and effort," he says, "it could be applied to literary texts as well. If or when this is successful, this will make an enormous difference to the way medievalists approach their subject.”
Register for the Launch (March 3)
Register for the Workshop (March 10)
Image credit: Christ Church MS Archives D&C vi.c.1, Special Collections at Christ Church, Oxford. © The Governing Body of Christ Church, Oxford