The Whole Is Bigger Than the Sum of Its Parts: Modeling Individual Annotators to Capture Emotional Variability

TL;DR

This paper introduces a novel approach to model individual annotators in emotion recognition, capturing nuanced emotional variability and improving the accuracy of emotion distributions over previous methods.

Contribution

It proposes a method to predict individual annotator labels and generate emotion distributions from continuous outputs, enhancing the modeling of subjective emotional data.

Findings

Emotion distributions are more accurate with the proposed method.

The approach outperforms prior work in within- and cross-corpus evaluations.

Captures inter-annotator variability effectively.

Abstract

Emotion expression and perception are nuanced, complex, and highly subjective processes. When multiple annotators label emotional data, the resulting labels contain high variability. Most speech emotion recognition tasks address this by averaging annotator labels as ground truth. However, this process omits the nuance of emotion and inter-annotator variability, which are important signals to capture. Previous work has attempted to learn distributions to capture emotion variability, but these methods also lose information about the individual annotators. We address these limitations by learning to predict individual annotators and by introducing a novel method to create distributions from continuous model outputs that permit the learning of emotion distributions during model training. We show that this combined approach can result in emotion distributions that are more accurate than those seen in prior work, in both within- and cross-corpus settings.