There are thousands upon thousands upon thousands of ancient documents from major civilizations from around the world that sit on shelves for wont of a translator. The number of people that can translate ancient text to modern languages is not that many, so the process of translating all these texts could take a century or more.
Not so fast, or maybe let’s go faster, say researchers from Macquaire University in Australia, who are working with artificial intelligence to train it to translate as much as possible with minimal human dependence. So far, the early results have produced promising results, though the ai still needs significant human aid to get it right.
How Do Archaeologists Crack the Code of Dead Languages?
From feedproxy.google.com
2021-09-10 19:00:00
Benjamin Plackett
Excerpt:
There are efforts to make translating ancient languages a more modern pursuit. Researchers at Macquarie University in Australia teamed up with experts from Google to use artificial intelligence with the aim of speeding up the process of translating ancient Egyptian hieroglyphs into English and Arabic.
“What the machine does well is to recognize where there are hieroglyphs and where there aren’t any. That’s not trivial because it means that the machine doesn’t just see gobbledygook,” says Camilla Di Biase-Dyson, a lecturer in Egyptology at Macquarie University who was involved in the project. “The problem is that it isn’t based on a great deal of training data. In order to snap a photo of an Egyptian tomb wall and translate it quickly, the machine will need a lot more data.”
For now, it still requires a lot of human input to make sure the end result is a reliable translation, but if the software is exposed to enough sample data, it’s possible that it may not need human help in the future.
About Author
