KuralHub lists 3 speech emotion recognition datasets for Polish, 1 of them openly available. Each entry below gives the emotion labels, size, recording setup, licence, where to get it and how to cite it. Polish is also part of the benchmark. The best of 11 models reached 87.0% test accuracy (whisper-large).
Polish 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 56.6% (median 66.0%). The best run reached 87.0% with whisper-large. Compare with other languages.
nEMO (run ID nEMO). Scores on the test split, best run highlighted.
The Polish Emotional Database consists of emotionally expressive speech recordings in Polish. It contains utterances spoken by professional actors who were instructed to express specific emotions. The database was created for research on emotion recognition systems and feature space derivation for affective speech processing. Each utterance was recorded to represent one of six emotional states: neutral, joy, anger, sadness, fear, and boredom.
Note: The dataset is not publicly available. You may need to contact the authors for access.
Citation
@inproceedings{cichosz2005low,
title={Low-dimensional feature space derivation for emotion recognition},
author={Cichosz, Jaroslaw and Slot, Krzysztof},
booktitle={Interspeech},
volume={5},
pages={477--480},
year={2005}
}
The Polish Multimodal Database of Emotional Speech, Video and Gestures is a comprehensive collection of emotional expressions in three modalities: facial expressions, body movements/gestures, and speech. Created by Sapinski et al. (2019), this dataset features recordings of 16 professional actors (8 males and 8 females, aged 25-64) from Teatr Nowy performing various emotional expressions.
The data was collected in a controlled environment - a quiet, well-lit theater rehearsal room with a green background. For facial expression recording, medium shots were used to keep the actor's face in frame while allowing for natural movement during emotional expressions. For body movement capture using Kinect, the recording frame included the entire body down to the legs.
This multimodal approach allows researchers to study emotional expressions across different communication channels simultaneously, making it a valuable resource for multimodal emotion recognition research, affective computing, and human-computer interaction studies focused on the Polish language.
Citation
@inproceedings{sapinski2019multimodal,
title={Multimodal database of emotional speech, video and gestures},
author={Sapi{\'n}ski, Tomasz and Kami{\'n}ska, Dorota and Pelikant, Adam and Ozcinar, Cagri and Avots, Egils and Anbarjafari, Gholamreza},
booktitle={Pattern Recognition and Information Forensics: ICPR 2018 International Workshops, CVAUI, IWCF, and MIPPSNA, Beijing, China, August 20-24, 2018, Revised Selected Papers 24},
pages={153--163},
year={2019},
organization={Springer}
}
The nEMO dataset is a Polish speech emotion corpus designed for Speech Emotion Recognition (SER) research. The corpus contains over 3 hours of emotional speech recordings performed by 9 professional actors. The dataset covers six basic emotions: anger, fear, happiness, sadness, surprise, and neutral.
The recordings are high-quality audio files suitable for developing and testing emotion recognition algorithms. The dataset was created by a Polish research team and has been made available for research purposes under a Creative Commons license.
Note: Use according to CC BY-NC-SA 4.0 license terms.
Used in the KuralHub benchmark (run ID nEMO). See the results above.
Citation
@inproceedings{christop-2024-nemo-dataset,
title = "n{EMO}: Dataset of Emotional Speech in {P}olish",
author = "Christop, Iwona",
editor = "Calzolari, Nicoletta and
Kan, Min-Yen and
Hoste, Veronique and
Lenci, Alessandro and
Sakti, Sakriani and
Xue, Nianwen",
booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",
month = may,
year = "2024",
address = "Torino, Italia",
publisher = "ELRA and ICCL",
url = "https://aclanthology.org/2024.lrec-main.1059",
pages = "12111--12116",
abstract = "Speech emotion recognition has become increasingly important in recent years due to its potential applications in healthcare, customer service, and personalization of dialogue systems. However, a major issue in this field is the lack of datasets that adequately represent basic emotional states across various language families. As datasets covering Slavic languages are rare, there is a need to address this research gap. This paper presents the development of nEMO, a novel corpus of emotional speech in Polish. The dataset comprises over 3 hours of samples recorded with the participation of nine actors portraying six emotional states: anger, fear, happiness, sadness, surprise, and a neutral state. The text material used was carefully selected to represent the phonetics of the Polish language adequately. The corpus is freely available under the terms of a Creative Commons license (CC BY-NC-SA 4.0).",
}
Know another Polish speech emotion dataset, or a newer version of one listed here? Suggest it or open a pull request.
This page is generated from datasets/Polish/ in the KuralHub repository.
If you use a dataset, cite its authors. If KuralHub helped you find it, please also cite the KuralHub paper.