AN EMPATHY MODEL OF TV AUDIENCE BY NEURAL NETWORKS USING DRAMA SCENARIO

Abstract

Researches on emotion have been done mainly in psychology. But in engineering, few studies on emotion have been reported. The reason is that emotion is very vague and depends on an individual. Recently, the emotions have been dealt with in kansei engineering, kansei information processing and emotional engineering and some human emotional models have been applied to industrial products. In this paper, we propose an empathy model by neural networks (NN). The empathy in this paper is one of emotions of an audience watching TV drama. First, we define eight emotion groups, 80 emotions and their emotion intensities. Secondly, we show an empathy model composed of eight NNs and training data with both “pure emotion intensities”and “empathy intensities”. Thirdly, after the NN finished the learning by the training data, we calculated empathies of hero's emotions extracted from the drama scenario book by the model and analyzed a story of the drama based on the empathies. The analysis was done successfully and we satisfied in the results. Finally, we compare the empathies by the model with human empathies by survey of the TV audience and show the usefulness of the model.