Analysing Gender Bias in IMDB Films Based on Social Networks

Abstract

The film industry has a major impact on society. This paper focuses on the gender bias in films. First, the social network analysis (SNA) method is used to construct a movie social network from the online movie dataset IMDb. By analysing these social network data, natural language processing methods are used to analyse movie titles, movie subtitles and casts. Then, this paper constructs a film gender bias classifier. Because the online movie data set IMDb lacks large-scale manual annotation data, and the classifier using traditional deep learning is prone to over-fitting, so this paper proposes a movie gender based on dual learning. The bias feature extractor and classifier enable training of the deep learning model from a small amount of annotated data and classify the bias. Finally, we find that women’s roles in the film have improved in all aspects, including the central role of female characters.