KuralHub lists 3 speech emotion recognition datasets for Spanish, 3 of them openly available. Each entry below gives the emotion labels, size, recording setup, licence, where to get it and how to cite it. Spanish is also part of the benchmark. The best of 11 models reached 68.0% test accuracy (whisper-small).
Spanish is one of the 29 languages in the KuralHub benchmark. Each of the 11 frozen speech models was trained with a small classification head on the shared emotions and scored on a held-out test split. Over all 11 runs the mean test accuracy was 36.5% (median 27.0%). The best run reached 68.0% with whisper-small. Compare with other languages.
MESD (run ID MESD). Scores on the test split, best run highlighted.
The Emotional Voice Messages (EMOVOME) database is a speech dataset collected for emotion recognition in real-world conditions. It contains 999 spontaneous voice messages from 100 Spanish speakers, collected from real conversations on a messaging app. EMOVOME includes both expert and non-expert emotional annotations, covering valence and arousal dimensions, along with emotion categories for the expert annotations. Detailed participant information is provided, including sociodemographic data and personality trait assessments using the NEO-FFI questionnaire. Moreover, EMOVOME provides audio recordings of participants reading a given text, as well as transcriptions of all 999 voice messages. Additionally, baseline models for valence and arousal recognition are provided, utilizing both speech and audio transcriptions.
Citation
@article{zaragoza2024emotional,
title={Emotional Voice Messages (EMOVOME) database: emotion recognition in spontaneous voice messages},
author={Zaragoz{\'a}, Luc{\'\i}a G{\'o}mez and del Amor, Roc{\'\i}o and Vargas, Elena Parra and Naranjo, Valery and Raya, Mariano Alca{\~n}iz and Mar{\'\i}n-Morales, Javier},
journal={arXiv preprint arXiv:2402.17496},
year={2024}
}
ESCorpus-PE dataset contains emotional utterances of Spanish peruvian speech gathered from Spanish interviews, TV reports, political debate and testimonials. It contains 3749 utterances of three emotional dimensions: Valence, Arousal and Dominance. There are 80 speakers (44 male and 36 female). This data was created from Youtube audios. These audios were selected following a specific criteria specified in the paper: ESCorpus-PE: A speech emotional database in Spanish with Peruvian accent, in the section Methods/Audio/Video Selection. Anyone can use this data only for research purposes.
Citation
@misc{cuno2021escorpus,
author = {Alvaro Cuno Parari and others},
title = {Escorpus-PE: A Speech Emotional Dataset in Spanish with Peruvian Accent},
year = {2021},
version = {1.0.0},
publisher = {Zenodo},
doi = {10.5281/zenodo.5793223},
url = {https://doi.org/10.5281/zenodo.5793223}
}
The Mexican Emotional Speech Database (MESD) contains single-word utterances expressing six emotions (anger, disgust, fear, happiness, neutral, sadness) with Mexican cultural shaping. It includes recordings from 3 female, 2 male, and 6 child non-professional actors, captured in a professional studio at 48kHz, 24-bit WAV format. The dataset features natural and naturalness-reduced versions, with modifications in pitch, formants, and intensity. Words were selected based on linguistic and emotional criteria. MESD is the first dataset of its kind for Mexican Spanish, covering both adult and child voices.
Used in the KuralHub benchmark (run ID MESD). See the results above.
Citation
@article{article,
author = {Duville, Mathilde Marie and Alonso-Valerdi, Luz and Ibarra-Zarate, David I.},
year = {2021},
month = {12},
pages = {},
title = {Mexican Emotional Speech Database Based on Semantic, Frequency, Familiarity, Concreteness, and Cultural Shaping of Affective Prosody},
journal = {Data},
doi = {10.3390/data6120130}
}
Know another Spanish speech emotion dataset, or a newer version of one listed here? Suggest it or open a pull request.
This page is generated from datasets/Spanish/ in the KuralHub repository.
If you use a dataset, cite its authors. If KuralHub helped you find it, please also cite the KuralHub paper.