"Inscriptions are an important source for ancient studies. The Romans alone left behind hundreds of thousands of them. But reading them and classifying them chronologically and, if necessary, geographically requires more than just a good knowledge of Latin, especially since many are only partially preserved."
A team led by London-based Google computer scientist Yannis Assael and classical philologist Thea Sommerschield from the University of Nottingham has now presented a computer system called "Aeneas" in the scientific journal "Nature." It uses a generative neural network to make suggestions for missing sections in Latin inscriptions.
The artificial intelligence also suggests dates and information on regional origins, making it easier to find relevant parallels.
The authors of the "Nature" study tested their software on a group of 23 Latin scholars with epigraphic expertise. In 90 percent of the cases in which they processed texts using the Aeneas system, the system's suggestions provided them with useful information. The scholars stated that they were given a starting point. However, the best results were generally achieved when the researchers worked on an inscription not alone or the software alone, but when both worked together.” [1]
1. KI hilft bei Latein. Frankfurter Allgemeine Zeitung; Frankfurt. 24 July 2025: 9. UvR
Komentarų nėra:
Rašyti komentarą