The researchers have turned to the Holy Book because they believe it contains “a large, previously untapped data set of aligned parallel text (or translation/s)”. Each version of the Bible reportedly contains over 31,000 verses that the researchers borrowed to produce more than 1.5 million unique pairs of source and translated texts. These pairs can then be fed into machine learning training sets.

“The English-language Bible comes in many different written styles, making it the perfect source text to work with for style translation,” commented Keith Carlson, a PhD student at Dartmouth. To make matters easier still, the Bible is already thoroughly indexed; it makes consistent use of book, chapter, and verse numbers, which reduces the number of potential errors with automatic translation and matching.