DialogueEIN: Emotional Interaction Network for Emotion Recognition in Conversations

Anonymous

Abstract: Emotion Recognition in Conversations (ERC) is a necessary step for developing empathetic human-computer interaction system. The existing methods on ERC primarily focus on capturing the context-level and speaker-level information from utterances. However, these methods ignore the causes of human emotion change, resulting in insufficient in capturing useful information for emotional prediction. In this work, we propose more explanatory Emotional Interaction Network (DialogueEIN) based on two main stages to capture the contexual information over intra- and inter-speaker dependencies directly from utterances, and to explore and analyze the differentiated contributions over the both kinds of information to boost better understanding of current utterance in conversation. Experimental results on two benchmark datasets demonstrate the effectiveness and superiority of our proposed model.

Paper Type: long