This Software Predicts A Book’s Popularity Based Only On Its Sentences
Statistical stylometry is the new literary critic.
Computer scientists at Stony Brook University in New York may have just put literary critics out of a job. The team developed a software program that predicts the likelihood of a book becoming a commercial success with 84% accuracy.
The predictive algorithm uses statistical stylometry, which mathematically assesses patterns and usage of grammar and vocabulary in books. The process analyzed thousands of downloadable books available in the Project Gutenberg archives, along with data from Amazon sales and national awards like the Pulitzer Prize. The study included books of every genre, from adventure and romance to science fiction and poetry, and determined that certain elements of writing style significantly contribute to the commercial success of a book.
Their findings would please your high school English teachers, who always preached you should “show, not tell” in your writing. The most successful books used many conjunctions (like “and” or “but”) and also had a higher frequency of nouns and adjectives, while less successful pieces relied more on verbs and adverbs that explicitly described emotions, like “wanted” or “promised.”
The team even broke down their findings to determine specific words that were more or less successful within each genre. Here are the results of their study of adventure stories:
So if that heart-pounding novel you’re working on features a risky love affair between slaves who live in the hills by the beach, perhaps you should rethink that storyline.