We propose a method of extracting TV program themes. Our approach detects burst words from TV programs and analyzes the time-series correlation between all pairs of them. In addition, we have developed a system for visualizing the appearance frequency of the keywords on the time axis for investigating the themes.
