Research on Emotion Prediction Based on Emotional Changes in Dialogue Scenarios

Le résumé fourni par la source

In recent years, the application of emotional analysis in dialogues in human-computer interaction has received increasing attention. The main difference between emotional analysis in dialogues and emotional analysis in single sentences is the contextual information that influences the emotional tone in dialogues. Previous studies have used deep learning to model context information and output the speaker’s current emotional state. In this study, we first proposed a sample angle selection mechanism to resolve the difficulty of uneven distribution of sample labels. Then, we designed a parallel convolutional neural network structure that uses both speech and text modal input signals to predict the emotional changes of the intervened (human) before and after being intervened by the external intervener (robot), and then predict their emotions. The proposed method was experimentally verified using the IEMOCAP dataset, and the results showed that the accuracy of the proposed method was improved compared to the results before incorporate the sample angle selection mechanism, demonstrating the effectiveness of the proposed method.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.