Deep learning for diplomatic video analysis

Abstract

In this paper, we focus on putting Faster-RCNN into practice to solve the problem of diplomatic video analysis, as the part of Mediated Public Diplomacy. Diplomatic video uploaded by U.S. Embassy in China is our research target. Using Faster-RCNN, we get 56,781 object detection results from those diplomatic videos. Then we use statistical tools to test the abnormal distribution of the object category “person”, clustering the above results so as to analyze the hidden strategic purposes in these diplomatic videos. Then we give an abstract of these videos: they mainly focus on common people’s high-quality life in the U.S. Strategic purposes are: the U.S. takes advantage of the repeating occurrence of common people to make “people to people” diplomacy in order to win hearts and minds of audience. Attractive personal life is depicted in the video so as to build a strong, harmonious and happy U.S. national image. These procedures are elaborately designed, which is a latent agenda setting process, and a fruitful frame construction attempt. By this way, the U.S. successfully bridge the culture gap and accomplish its goal of global hegemony.