Staff at the University of Notre Dame used AI to analyze ancient manuscripts.
Experts said that there is a huge amount of materials and manuscripts that have survived to this day, but they have not been read.
Researchers at the University of Notre Dame are now developing AI to read complex ancient handwriting in order to improve its transcription.
We decided to automate the transcription in such a way that it mimics the way an experienced reader perceives the text.
Walter Scheirer, assistant professor at Dennis O. Doughty College
In a new work, the authors combined traditional machine learning methods with visual psychophysics – a method of measuring connections between physical stimuli and mental phenomena, such as the amount of time it takes for an experienced reader to recognize a particular symbol, handwriting quality, or determine how an abbreviation stands.
Shairer’s team examined digitized Latin manuscripts that were written by scribes in the monastery of Saint Gall in the ninth century. Readers entered transcriptions into a specially designed software interface. The team then measured reaction times during transcription to understand which words, symbols, and passages were easy and which were difficult. Scheirer explained that this is necessary to get closer to human perception of the text, so the number of errors will decrease and AI will decipher the text more accurately.
The team is currently working on improving the accuracy of the transcriptions, especially when working with a damaged or incomplete document, as well as on how to perceive the image and other accompanying symbols that can lead to confusion.