Evaluating Gender Bias in Film Dialogue

Lesley Istead, Andreea Pocol, Sherman Siu

Abstract: In this paper, we estimate gender bias in movies by evaluating the use of ‘gendered language,’ or utterances pertaining to a particular gender, such as he, she, mother, and father. We compute the bias over a corpus of over 40 thousand movie subtitle files and explore changes in bias by release year, genre, country of origin, MPAA rating, IMDb scores, and box office records. Our analysis indicates a prevalent male bias, with more than 75% of recent films exhibiting this bias. However, we observed that this bias has been decreasing over the last fifteen years. We also note that the disparity in box office sales between films with male and female bias has been closing .