Quantitative analysis of the evolution of novelty in cinema through crowdsourced keywords

TL;DR

This study quantitatively analyzes the evolution of novelty in American cinema over 70 years using crowdsourced keywords, revealing how novelty impacts revenue and its statistical properties.

Contribution

It introduces a large-scale, data-driven method to measure film novelty through keyword analysis and explores its relationship with revenue and content dynamics.

Findings

Novelty influences film revenue following a Wundt-Berlyne-like curve.

Keyword occurrence statistics reveal content evolution over a century.

Quantitative measures of novelty provide new insights into cinematic trends.

Abstract

The generation of novelty is central to any creative endeavor. Novelty generation and the relationship between novelty and individual hedonic value have long been subjects of study in social psychology. However, few studies have utilized large-scale datasets to quantitatively investigate these issues. Here we consider the domain of American cinema and explore these questions using a database of films spanning a 70 year period. We use crowdsourced keywords from the Internet Movie Database as a window into the contents of films, and prescribe novelty scores for each film based on occurrence probabilities of individual keywords and keyword-pairs. These scores provide revealing insights into the dynamics of novelty in cinema. We investigate how novelty influences the revenue generated by a film, and find a relationship that resembles the Wundt-Berlyne curve. We also study the statistics of keyword occurrence and the aggregate distribution of keywords over a 100 year period.