KuralHub lists 2 speech emotion recognition datasets for Kazakh. Each entry below gives the emotion labels, size, recording setup, licence, where to get it and how to cite it. Kazakh is also part of the benchmark. The best of 11 models reached 65.0% test accuracy (whisper-small).
Kazakh 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 44.0% (median 51.0%). The best run reached 65.0% with whisper-small. Compare with other languages.
KazEmoTTS (run ID KazakhEmotionalTTS). Scores on the test split, best run highlighted.
KazEmoTTS is a specialized dataset designed for Kazakh emotional text-to-speech (TTS) synthesis. It comprises 54,760 audio-text pairs totaling 74.85 hours of speech, recorded by three narrators (one female and two male) covering six emotions: "neutral," "angry," "happy," "sad," "scared," and "surprised" 148. The dataset is structured to include diverse vocal expressions, with the female narrator contributing 34.23 hours and the male narrators 40.62 hours combined
Used in the KuralHub benchmark (run ID KazakhEmotionalTTS). See the results above.
Citation
@article{abilbekov2024kazemotts,
title={KazEmoTTS: A Dataset for Kazakh Emotional Text-to-Speech Synthesis},
author={Abilbekov, Adal and Mussakhojayeva, Saida and Yeshpanov, Rustem and Varol, Huseyin Atakan},
journal={arXiv preprint arXiv:2404.01033},
year={2024}
}
The Kazakh-Russian Emotion Corpus is a multilingual dataset designed fmor speech emotion recognition (SER) in both Kazakh and Russian languages. It contains over 16,000 audio recordings labeled with 8 emotion categories, collected from 101 participants to ensure diversity in vocal expressions. The dataset is divided into training (80%), validation (10%), and testing (10%) sets, facilitating robust model evaluation and development for SER tasks.
Citation
@article{kozhakhmet2020speech,
title={Speech emotion recognition for Kazakh and Russian languages},
author={Kozhakhmet, Kanat and Zhumaliyeva, Rakhima and Shoinbek, A and Sultanova, Nazerke},
journal={Appl. Math. Inf. Sci},
volume={14},
pages={65--68},
year={2020}
}
Know another Kazakh speech emotion dataset, or a newer version of one listed here? Suggest it or open a pull request.
This page is generated from datasets/Kazakh/ in the KuralHub repository.
If you use a dataset, cite its authors. If KuralHub helped you find it, please also cite the KuralHub paper.