Long-Short Distance Graph Neural Networks and Improved Curriculum Learning for Emotion Recognition in Conversation

TL;DR

This paper introduces a novel multimodal graph neural network architecture, LSDGNN, combined with an improved curriculum learning strategy, to enhance emotion recognition in conversation by capturing long- and short-distance features and addressing data imbalance.

Contribution

It proposes LSDGNN, a new multimodal graph neural network leveraging DAGs for long- and short-distance features, and an improved curriculum learning method for better emotion recognition.

Findings

Outperforms existing benchmarks on IEMOCAP and MELD datasets

Effectively captures long- and short-distance utterance features

Addresses data imbalance with a novel curriculum learning approach

Abstract

Emotion Recognition in Conversation (ERC) is a practical and challenging task. This paper proposes a novel multimodal approach, the Long-Short Distance Graph Neural Network (LSDGNN). Based on the Directed Acyclic Graph (DAG), it constructs a long-distance graph neural network and a short-distance graph neural network to obtain multimodal features of distant and nearby utterances, respectively. To ensure that long- and short-distance features are as distinct as possible in representation while enabling mutual influence between the two modules, we employ a Differential Regularizer and incorporate a BiAffine Module to facilitate feature interaction. In addition, we propose an Improved Curriculum Learning (ICL) to address the challenge of data imbalance. By computing the similarity between different emotions to emphasize the shifts in similar emotions, we design a "weighted emotional shift" metric and develop a difficulty measurer, enabling a training process that prioritizes learning easy samples before harder ones. Experimental results on the IEMOCAP and MELD datasets demonstrate that our model outperforms existing benchmarks.