EmoGraph: Capturing Emotion Correlations using Graph Networks

TL;DR

This paper introduces EmoGraph, a graph network-based approach that models emotion correlations to improve multi-label and single-label emotion classification tasks, demonstrating superior performance over existing methods.

Contribution

EmoGraph is the first to explicitly model emotion interdependencies using graph networks based on co-occurrence statistics, enhancing classification accuracy.

Findings

Outperforms strong baselines on multi-label datasets

Improves macro-F1 scores significantly

Benefits single-label classification tasks

Abstract

Most emotion recognition methods tackle the emotion understanding task by considering individual emotion independently while ignoring their fuzziness nature and the interconnections among them. In this paper, we explore how emotion correlations can be captured and help different classification tasks. We propose EmoGraph that captures the dependencies among different emotions through graph networks. These graphs are constructed by leveraging the co-occurrence statistics among different emotion categories. Empirical results on two multi-label classification datasets demonstrate that EmoGraph outperforms strong baselines, especially for macro-F1. An additional experiment illustrates the captured emotion correlations can also benefit a single-label classification task.