Emotion Recognition using Sequence Mining

Abstract

With the development of human-computer interaction technology, user emotion recognition, as an important factor in the process of natural language communication, has become a hot research topic. Current studies mainly analyze the emotional within a single long sentence, but in real communication, the transmission of information and emotion is more often achieved by multi-round dialogue. In this paper, we construct the emotional sequence between people and people in different scenarios to identify their multi-round conversational emotions, and analyze their emotional changes through sequence mining. This article takes several novels as the analysis corpus, and proceeds the scene segmentation according to the chapters of the novel. Then we analyze the emotional bias of each conversation between different people in each scene, and then construct the emotional matrix between people in each scene. Finally, the LSTM algorithm is used to mine different emotional patterns and changing trends between people compared with machine learning algorithm. Experimental results show that the proposed sequential-based emotion recognition method can recognize the emotions between people very well and predict the future emotional patterns.