Home to some 35, 000 volumes and with over 80 km of bookshelf-filled shelves, the Vatican Secret Archives are home to one of humanity's most important historical collections, with the oldest manuscript dating from the late 8th century. The problem is that it cannot be accessed virtually, but is only available to academics who can move to the headquarters of the Catholic Church.

A group of researchers from Roma Tre University and La Sapienza University, both in the capital of Italy, want to change that. They are developing a project called In Codice Ratio, which aims to automatically transcribe all the contents of the Archives that can be accessed by the public.

To this end, the team works with a combination of artificial intelligence and optical character recognition (OCR). They use a new technique to be able to apply current OCR knowledge to handwritten texts by dividing the words into small pieces that depend on the amount of ink used in the region. This allows you to separate each letter individually and turn the paper into a digital document.

On the left, the handwritten text; in the middle, the right word; and on the right, what the program has identified.

The artificial intelligence training responsible for recognizing these pieces of text as letters of the Latin alphabet was done by Italian high school students. They answered several questionnaires and identified which images represented the desired letters.

The result of all this was put to the test when the team performed the tests using about 18, 000 pages of letters stored in the Archives. According to the researchers, the program was able to correctly identify 96% of the letters, with one third of the words containing at least one wrong character. Because this technology always learns when its errors are identified, these numbers are only likely to improve on future attempts.

On the In Codice Ray website, you can see all articles published with the search results.

Researchers use AI to scan Vatican Secret Files via TecMundo