Audio-visual and language measures

The following projects describe the different measures we have used and developed for media content analyses

Movie on-screen time

Predicting race from face

Gender from audio

Multimodal and interaction analysis

Modeling multiple modalities and analyzing character interactions provide a complementary representation of movies and characters

Multimodal autoencoders

Connecting content measures to impact

We investigate how the measures we develop are related to predicting impact of content

Violence Prediction from Movie Scripts

Knowledge transfer to predict affect