CAMEO: Collection of Multilingual Emotional Speech Corpora

TL;DR

This work addresses the lack of standardized, multilingual data for speech emotion recognition by introducing CAMEO, a curated collection of 13 datasets across eight languages released on Hugging Face. It details systematic data curation, consistent metadata, and an evaluation framework with a public leaderboard to enable reproducible cross-language benchmarking. The study demonstrates the difficulty of multilingual zero-shot SER, reporting modest macro F1 scores across models and highlighting potential data overlap with well-known benchmarks. By providing open-source tooling and a continuously updating leaderboard, CAMEO aims to accelerate robust cross-lingual SER research and fair model comparisons.

Abstract

This paper presents CAMEO -- a curated collection of multilingual emotional speech datasets designed to facilitate research in emotion recognition and other speech-related tasks. The main objectives were to ensure easy access to the data, to allow reproducibility of the results, and to provide a standardized benchmark for evaluating speech emotion recognition (SER) systems across different emotional states and languages. The paper describes the dataset selection criteria, the curation and normalization process, and provides performance results for several models. The collection, along with metadata, and a leaderboard, is publicly available via the Hugging Face platform.

Figures (1)

• Figure 1: Example illustrating the application of the post-processing strategy used to extract labels from generated responses.