MMGCN: Multimodal Fusion via Deep Graph Convolution Network for Emotion Recognition in Conversation

TL;DR

This paper introduces MMGCN, a novel multimodal graph convolutional network that effectively models multimodal and speaker dependencies for emotion recognition in conversations, outperforming existing methods on benchmark datasets.

Contribution

The paper presents a new multimodal fused graph convolutional network that leverages both multimodal and speaker information for improved emotion recognition in conversations.

Findings

MMGCN outperforms state-of-the-art methods on IEMOCAP and MELD datasets.

Effective modeling of multimodal dependencies improves emotion recognition accuracy.

Utilizing speaker information enhances inter- and intra-speaker dependency modeling.

Abstract

Emotion recognition in conversation (ERC) is a crucial component in affective dialogue systems, which helps the system understand users' emotions and generate empathetic responses. However, most works focus on modeling speaker and contextual information primarily on the textual modality or simply leveraging multimodal information through feature concatenation. In order to explore a more effective way of utilizing both multimodal and long-distance contextual information, we propose a new model based on multimodal fused graph convolutional network, MMGCN, in this work. MMGCN can not only make use of multimodal dependencies effectively, but also leverage speaker information to model inter-speaker and intra-speaker dependency. We evaluate our proposed model on two public benchmark datasets, IEMOCAP and MELD, and the results prove the effectiveness of MMGCN, which outperforms other SOTA methods by a significant margin under the multimodal conversation setting.