Predicting Romantic Comedy Success From Content

Abstract

Predicting film success has proven challenging, with prior research examining factors including budget, production studios, and stars, to varied degrees of accuracy. Missing was the impact of film scripts and latent linguistic features, examined here through textual analysis. Recent computational study identified the latent thematic content of nearly 200 romantic comedy films, revealing an increasing focus on romantic relationships and tumultuous courtship. We harness the same linguistic model to test whether changes in thematic content are associated with success in reviews, awards, and financial earnings. We find relationship-centered content is positively associated with earnings, mediated by the number of theaters.