An analysis of gendered language patterns in film scripts using natural language processing and log odds ratios to identify how male and female characters are portrayed differently through action words and descriptions. Inspiration: This project was inspired by The Pudding's "She Giggles, He Gallops" article, which analyzed gender tropes across 2,000 film scripts.
This analysis examines bigrams (two-word sequences) in film scripts to discover which action words are more strongly associated with "she" vs "he" in screen directions. For example:
Words associated with "she": giggles, sobs, smiles, gasps Words associated with "he": shoots, gallops, grins, kills
The analysis uses log odds ratios to quantify these associations, revealing underlying gender stereotypes in how screenwriters portray male and female characters.
Bigram Extraction: Identifies pronoun + action word pairs (e.g., "she smiles", "he runs") Log Odds Ratio Analysis: Quantifies which words are more strongly associated with each gender Statistical Smoothing: Implements Laplace smoothing to handle rare words Visualization: Creates bar charts showing the most gendered words Stop Word Filtering: Removes common words to focus on meaningful actions
Credits Original Research: Julia Silge, Russell Samora, Amber Thomas, Hanah Anderson (The Pudding)