Mysterious Voynich manuscript deciphered using AI
Researchers claim to have decoded one of the most mysterious documents ever found, using a series of algorithms to translate the enigmatic Voynich manuscript.
The illustrated codex, named after the Polish book dealer who purchased it in 1912, has stumped generations of scholars with its impenetrable text and drawings of alien-looking plants. While the document has been dated to the 15th century, what it actually says has been the source of debate. Is it written in an abbreviated form of Latin? Is it a guide to herbal remedies? Is it an alchemy textbook? Is it gibberish?
Grzegory Kondrak and Bradley Hauer from the University of Alberta may have cracked a key part of the Voynich code, with the help of a language-deciphering artificial intelligence. In their published account of the project, the computer scientists explain how they fed samples from 380 languages into an algorithmic system, which was then set on the opaque text of the medieval codex.
Their findings suggest that the Voynich manuscript may be written in a form of Hebrew, encoded using alphabetically ordered anagrams of the individual words. Kondrak commented that the result was “surprising”, and acknowledged that more work is needed to properly decode what the manuscript actually means: “Just saying ‘this is Hebrew’ is the first step. The next step is how do we decipher it.”
The language recognition system found that 80% of the words in the manuscript were in a Hebrew dictionary. The codex contains words such as “light”, “air” and “farmer”, although actually making comprehensible sentences from these fragments is a difficult task. It was only by working with experts in ancient Hebrew that the team was able to glean something approaching normal syntax. According to the results, the first sentence of the text reads: “She made recommendations to the priest, man of the house and me and people.”
It’s worth noting that an attempt was made last year by historian Nicholas Gibbs to translate the Voynich manuscript as a woman’s health manual, based on an abbreviated form of Latin. Gibbs’ claim was swiftly pounced on by experts that poked holes in his hypothesis, saying the results were not grammatically correct. Perhaps Kondrak and Hauer will have better luck from the research community, or maybe they will face similar criticism.
Further collaboration with historians will be needed to make more sense from the seemingly jumbled Hebrew, but the latest piece of research demonstrates the scope for algorithmic translation to work alongside human experts. Kondrak said he would be keen to apply the software to other ancient texts. Perhaps a good place to start would be this freely available digital library of 3,500 occult manuscripts, established by Amsterdam’s Ritman Library.