Applying domain-specific natural language understanding techniques to film

Abstract

We report on our work involving adapting Natural Language Processing (NLP) tools, including subject-predicate-object triplet generation, to build a system capable of automatically testing if a given film or play passes the Bechdel-Wallace test, a test from feminist film theory. This test is passed when a film contains two or more females who discuss a topic other than a man. Despite efforts by researchers, no system or algorithm has yet been developed that can definitively state whether or a not a film has passed the test. Our system uses tools that have not yet been applied to this problem, thus expanding the body of knowledge and making new progress towards this goal.