Friday, January 10, 2014

RESEARCHERS FIND AN ALGORITHM TO PREDICT BOOK SUCCESS, LIKE A MAGIC 8 BALL FOR FICTION SALES

(from bustle.com
by Claire Luchette)

“Will my book be a hit?” Worry not: researchers in the computer science department at Stony Brook University can predict whether a book will sell with 84 percent accuracy. They’ve developed an algorithm that compares a writer’s style with that of bestselling and low-selling authors.

The method of comparison, called statistical stylometry, involves studing the linguistic style in the first 1000 sentences of an author’s work: quantifying the parts of speech, lexical choices, and connotative categories in an author’s first 1,000 sentences. The study analyzed bestselling works across the genres of Adventure, Detective/Mystery, Fiction, Historical Fiction, Love Stories, Poetry, Science Fiction, and Short Stories. (It’s not clear from the research how the researchers sorted cross-genre works.)

While previous studies have analyzed the actual literary content — plot, characteristics of characters, relationships — or the qualities of significant authors’s styles, this is the first study that’s quantitative in nature. “To our knowledge, no prior work has undertaken a systematic quantitative investigation on the overarching characterization of the writing style in successful literature,” the researchers wrote.

The researchers found that the authors of successful novels used a higher number of conjunctions, prepositions, nouns, pronouns, determiners and adjectives. The language of successful books also refers to thought-processing — words like “recognized” and “remembered.”

Authors of less successful books used a higher percentage of verbs — particularly explicit action verbs like “wanted” and “cheered” — as well as adverbs, and foreign words. Less successful books also have more cliches and negative words.

So the formula is easy: titrate some nouns with some prepositions, stir in adjectives, and filter out the action verbs. Bake for six years and call an agent.