Artificial Intelligence Takes a Crack at Decoding the Mysterious Voynich Manuscript
But medieval scholars are skeptical about this latest attempt to decipher the world’s “most mysterious book”
The Voynich Manuscript has baffled cryptographers ever since the early 15th-century document was rediscovered by a Polish book dealer in 1912. The handwritten, 240-page screed, now housed in Yale University's Beinecke Rare Book & Manuscript Library, is written from left to right in an unknown language. On top of that, the text itself is likely to have been scrambled by an unknown code. Despite numerous attempts to crack the code by some of the world’s best cryptographers, including Alan Turing and the Bletchley Park team, the contents of the enigmatic book have long remained a mystery. But that hasn't stopped people from trying. The latest to give it a stab? The Artificial Intelligence Lab at the University of Alberta.
Bob Weber at the Canadian Press reports that natural language processing expert Greg Kondrak and grad student Bradley Hauer have attempted to identify the language the manuscript was written in using AI. According to a press release, the team originally believed that the manuscript was written in Arabic. But after feeding it to an AI trained to recognize 380 languages with 97 percent accuracy, its analysis of the letter frequency suggested the text was likely written in Hebrew.
“That was surprising,” Kondrak says. They then hypothesized that the words were alphagrams, in which the letters are shuffled and vowels are dropped. When they unscrambled the first line of text using that method they found that 80 percent of the words created were found in the Hebrew dictionary. The research appears in the journal Transactions of the Association of Computational Linguistics.
Neither of the researchers are schooled in ancient Hebrew, so George Dvorsky at Gizmodo reports they took their deciphered first line to computer scientist Moshe Koppel, a colleague and native Hebrew speaker. He said it didn’t form a coherent sentence. After the team fixed some funky spelling errors and ran it through Google Translate, they came up with something readable, even if it doesn’t make much sense: “She made recommendations to the priest, man of the house and me and people.”
Weber reports that a translation of another 72-word section included the words "farmer," "light," "air" and "fire."
"Somebody with very good knowledge of Hebrew and who’s a historian at the same time could take this evidence and follow this kind of clue,” Kondrak tells Weber.
But Voynich scholars are skeptical. Medievalist Damian Fleming of Indiana University–Purdue University Fort Wayne was among those who responded to news of the work in frustration on social media, specifically critiquing the decision to use Google Translate to decipher the manuscript rather than consult a Hebrew scholar.
*howling in voynich language* pic.twitter.com/ObNXNpVTVy
— Damian ᚠלεᛗιnג (@FW_Medieval) January 30, 2018
Experts familiar with the Voynich manuscript and its history have good reason to express doubt; over the years many people have made numerous claims that they've deciphered the manuscript, only to be proven wrong. One of the latest cases happened last year when an amateur code-breaker Nicholas Gibbs claimed he’d figured out that the book was written primarily in Latin abbreviation. After Gibb's declaration, medievalists came out of the woodwork to dispute his findings, saying the abbreviation theory did not add up to grammatically correct Latin and did not make sense. It also appears that Gibbs came up with the theory to try and sell a show about the manuscript to a cable network.
Though we still do not know what the book says, researchers have several hypotheses about what the manuscript is about. Based on the book's illustrations of plants and bathing women, a number of scholars believe that it's actually a medical textbook about women’s health—a subject so mysterious that it had to be hidden away in one of the world's most perplexing manuscripts.