Detecting Gender Inequality and Language Evolution in Movie Dialogue

Abstract

It is expected that gender inequality in American movies has decreased over the last century, reflecting larger societal trends. But, can this be verified through a computational analysis of movie dialogue? In this study, we apply 13 word-based metrics intended to rate words according to such criteria as gender-ladenness, pleasantness, and age of acquisition, and ask whether any are effective at measuring gender differences in movie dialogue. As a secondary outcome, we also seek to determine whether gender differences have decreased over time and how. We find that metrics that aim to capture word concreteness, imageability, pleasantness, easiness of understanding, and gender-ladenness are able to detect gender differences in dialogue. We also find evidence that at least since the 1960s, there has been a steady convergence in female and male dialogue in Hollywood films.