If only robots could have prevented Dark Phoenix.
As legendary screenwriter William Goldman once said about making a successful movie, “nobody knows nothing.” But a new AI bot is trying where humans have failed. The bot’s builders claims the program can determine a movie’s critical and financial success simply by reading a written description of the plot.
“As the size of investment for movie production grows bigger, the need for predicting a movie’s success in early stages has increased,” say study authors You Jin Kim, Yun Gyung Cheong, and Jung Hoon Lee in their paper’s abstract. “To address this need, various approaches have been proposed, mostly relying on movie reviews, trailer movie clips, and [social media] postings. However, all of these are available only after a movie is produced and released.”
So the researchers examined a dataset of 42,306 movies from all over the world. They took their plot summaries from Wikipedia and their critical rankings from Rotten Tomatoes, both of which might be up to bias in one way or another. Such are the risks of attempting to feed art into an algorithm.
On each level, the researchers divided the movies into 0 and 1. Flops were assigned a 0, while hits were assigned a 1. The same goes for criticism: Panned movies were given a 0, and acclaimed movies given a 1.
“Since these scores indicate the popularity and quality of a movie, we define a successful movie as having the combination of these score greater than 75,” the researchers say in their paper. As an example, the first Avengers film, from 2012, scores a .78. Films like Alice in Wonderland, Das Boot, and A Man for All Seasons scored similarly high.
The researchers suggest that a film’s overall tone doesn’t affect its success, but that “the frequency of sentiment changes may signal the success of films.” Making sure that a movie has more than one emotional beat isn’t exactly a revelation in terms of filmmaking, but it is interesting to see the data reveal that emotional truths could be right all along.
The researchers presented their paper at the Storytelling Workshop 2019 in Florence, Italy.
This article was written by David Grossman and published by Popular Mechanics on 2/08/2019.