In 2017, a project based in Roma Tre University called In Codice Ratio began using artificial intelligence and optical character recognition to attempt to transcribe more documents from the archives.[3][5] While character-recognition software is adept at reading typed text, the cramped and many-serifed style of medieval handwriting makes distinguishing individual characters difficult for the software.[6] Many individual letters of the alphabet are often confused by human readers of medieval handwriting, let alone a computer program. The team behind In Codice Ratio tried to solve this problem by developing a machine-learning software that could parse this handwriting. Their program eventually achieved 96% accuracy in parsing this type of text.[7]