KuralHub lists 3 speech emotion recognition datasets for Greek, 2 of them openly available. Each entry below gives the emotion labels, size, recording setup, licence, where to get it and how to cite it. Greek is also part of the benchmark. The best of 11 models reached 75.0% test accuracy (hubert-base-ls960).
Greek 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 46.6% (median 54.0%). The best run reached 75.0% with hubert-base-ls960. Compare with other languages.
AESDD (run ID AESDD). Scores on the test split, best run highlighted.
AESDD v1.0 was created on October 2017 in the Laboratory of Electronic Media, School of Journalism and Mass Communications, Aristotle University of Thessaloniki, for the needs of Speech Emotion Recognition research of the Multidisciplinary Media & Mediated Communication Research Group (M3C, http://m3c.web.auth.gr/). It is a collection of utterances of emotional speech acted by professional actors. This version is the initial state of AESDD. The purpose of this project the continuous growth of the database through the collaborative effort of the M3C research group and theatrical teams.
Used in the KuralHub benchmark (run ID AESDD). See the results above.
Citation
@inproceedings{vryzas2018speech,
title={Speech emotion recognition adapted to multimodal semantic repositories},
author={Vryzas, Nikolaos and Vrysis, Lazaros and Kotsakis, Rigas and Dimoulas, Charalampos},
booktitle={2018 13th International Workshop on Semantic and Social Media Adaptation and Personalization (SMAP)},
pages={31--35},
year={2018},
organization={IEEE}
}
@article{vryzas2018speech,
title={Speech emotion recognition for performance interaction},
author={Vryzas, Nikolaos and Kotsakis, Rigas and Liatsou, Aikaterini and Dimoulas, Charalampos A and Kalliris, George},
journal={Journal of the Audio Engineering Society},
volume={66},
number={6},
pages={457--467},
year={2018},
publisher={Audio Engineering Society}
}
The GreThE (Greek Theatrical Emotion Dataset) is a collection of 500 audio samples from 90 unique speakers across 23 Greek theatrical plays. The dataset is annotated for valence(negative/neutral/positive) and arousal (weak/neutral/strong). It is designed for Speech Emotion Recognition (SER) tasks, particularly in the context of theatrical performances. The dataset includes Mel-spectrograms and segment feature vectors for each audio sample, making it suitable for machine learning and deep learning experiments.
Citation
@article{moutti2022dataset,
title={A Dataset for Speech Emotion Recognition in Greek Theatrical Plays},
author={Moutti, Maria and Eleftheriou, Sofia and Koromilas, Panagiotis and Giannakopoulos, Theodoros},
journal={arXiv preprint arXiv:2203.15568},
year={2022}
}
The SEWA dataset is a multimodal dataset designed for emotion recognition and analysis. It includes audio and video data collected from participants during natural interactions. The dataset aims to study emotional states in real-world scenarios, focusing on cross-cultural differences and spontaneous emotional expressions. Data was collected using wearable devices and audio-visual recordings, making it a rich resource for affective computing research.
Citation
@article{kossaifi2019sewa,
title={Sewa db: A rich database for audio-visual emotion and sentiment research in the wild},
author={Kossaifi, Jean and Walecki, Robert and Panagakis, Yannis and Shen, Jie and Schmitt, Maximilian and Ringeval, Fabien and Han, Jing and Pandit, Vedhas and Toisoul, Antoine and Schuller, Bj{\"o}rn and others},
journal={IEEE transactions on pattern analysis and machine intelligence},
volume={43},
number={3},
pages={1022--1040},
year={2019},
publisher={IEEE}
}
Know another Greek speech emotion dataset, or a newer version of one listed here? Suggest it or open a pull request.
This page is generated from datasets/Greek/ in the KuralHub repository.
If you use a dataset, cite its authors. If KuralHub helped you find it, please also cite the KuralHub paper.