It might not surprise you to know that people love to watch animal videos. But did you know that people are more likely to watch if the animal is splashing around in water?
This is the kind of information you can mine and analyze from audience behavior data gathered from TV ratings and multimedia content.
Researchers are keenly interested in what captures and loses a viewer’s attention because this information can be used to create better television programming, produce higher ratings, and attract more sponsors, say authors Ryota Hinami of University of Tokyo and Shin’ichi Satoh of National Institute of Informatics.
“To discover relationships between TV ratings and multimedia content, we focus on the change points—that is, the points in time when people first tune in to a particular TV program. Because these points reflect the active intention of TV viewers, they contain valuable information about viewers’ interests. We describe these points using visual features extracted from video and using keywords extracted from transcripts,” they write.