Canadian scientists used machine learning to decipher the Voynich manuscript, a famous text written by an unknown author in an unknown language. The authors claim that they were able to identify the language of the manuscript - it turned out to be Hebrew - and to translate the first sentence. The results achieved were reported by the Canadian channel CTV News.
The Voynich manuscript is a medieval 240-page codex in an unknown language, accompanied by a large number of illustrations, including unknown plants, rituals of incomprehensible purpose and complex diagrams, presumably related to astrology. According to the results of radiocarbon analysis, the paper on which it is written was made in the early 15th century. There are many assumptions about the meaning of the text, but none of them is accepted by the entire scientific community. Some scholars argue that the entire manuscript is a hoax.
Greg Kondrak of the University of Alberta trained a neural network to translate the Universal Declaration of Human Rights into 380 languages. As a result, the program could correctly detect the language 97% of the time. According to the results of the neural network, the first sentence of the manuscript is translated from Hebrew as “She made recommendations to the priest, the man at home, and me and people” (“She made recommendations to the priest, man of the house and me and people”). It was also concluded that the letters in the words are reversed, and the vowels are missing.
In one of the sections, words were found suitable for a botanical text: farmer, light, air and fire. Kondrak admits that other Voynich manuscript specialists accepted his work without enthusiasm. “I don't think they are supportive of this kind of research,” says the author. "People may be afraid that computers will replace them."