EMO-SUPERB: An In-depth Look at Speech Emotion Recognition

TL;DR

This paper introduces EMO-SUPERB, a comprehensive benchmark for speech emotion recognition that evaluates multiple models across datasets, promotes open collaboration, and explores using ChatGPT for annotation enhancement.

Contribution

The paper presents EMO-SUPERB, a new open-source benchmark with a unified evaluation framework and an innovative approach using ChatGPT to improve annotation quality in SER.

Findings

EMO-SUPERB includes 15 SSL models and 6 datasets for comprehensive evaluation.

Using ChatGPT for annotation re-labeling yields a 3.08% average relative gain.

The benchmark fosters community collaboration through an online leaderboard.

Abstract

Speech emotion recognition (SER) is a pivotal technology for human-computer interaction systems. However, 80.77% of SER papers yield results that cannot be reproduced. We develop EMO-SUPERB, short for EMOtion Speech Universal PERformance Benchmark, which aims to enhance open-source initiatives for SER. EMO-SUPERB includes a user-friendly codebase to leverage 15 state-of-the-art speech self-supervised learning models (SSLMs) for exhaustive evaluation across six open-source SER datasets. EMO-SUPERB streamlines result sharing via an online leaderboard, fostering collaboration within a community-driven benchmark and thereby enhancing the development of SER. On average, 2.58% of annotations are annotated using natural language. SER relies on classification models and is unable to process natural languages, leading to the discarding of these valuable annotations. We prompt ChatGPT to mimic annotators, comprehend natural language annotations, and subsequently re-label the data. By utilizing labels generated by ChatGPT, we consistently achieve an average relative gain of 3.08% across all settings.