The book consists of around 240 pages of inscrutable illustrations and coded language. It even has fold out pages, which is incredible for a book written some 600 years ago. It’s been owned by all sorts of people, from alchemists to emperors, since it was written in the 15th century. It finally got it’s modern name in the 20th century after a Polish book dealer, Wilfrid Voynich.
All sorts of codebreakers, cryptographers and folks from other walks of life have tried making sense of the seemingly undecipherable language, but to no avail. Until now.
A few Canadian data scientists may finally have broken a part of the code. Using a technique called ‘algorithmic decipherment’, these researchers from the University of Alberta have used artificial intelligence to crack the code hidden in this ancient text.
The algorithm was tested on 380 different translations and was able to recognise the language of origin (by looking for patterns) an impressive 97% of the time. This model was then applied on the Voynich manuscript. The researchers suspected it was written in Arabic, but the model disagreed. It indicated Hebrew as the most likely language.
Following this, they trained the model to look for alphagram (a mixture of alphabetically organised anagrams) patterns in the text. As it turned out, almost 80% of the words in the book were present in the Hebrew dictionary. But the issue still plaguing the model is that when put together, the sentences don’t often make much sense.
The research is still continuing and experts in the Hebrew language have been called in to look at the manuscript as well. You can read the research paper here.
Our take on this
This is an exciting time in the world of Natural Language Processing. No doubt researchers all over the world will be clamouring to apply (and improve) this algorithm to shine a new light on more ancient texts that have eluded historians so far.
You can also read this article on our Mobile APP