KuralHub lists 3 speech emotion recognition datasets for Tamil, 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. Tamil is also part of the benchmark. The best of 11 models reached 44.0% test accuracy (whisper-small).
Tamil 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 30.6% (median 29.0%). The best run reached 44.0% with whisper-small. Compare with other languages.
EmoTa (run ID EmoTa). Scores on the test split, best run highlighted.
Collected by the Emota team, this dataset contributes to Tamil Speech Emotion Recognition (SER) focusing on Sri Lankan Tamil. It includes recordings from undergraduates who took Drama Grade B or above at the O/L level, covering diverse regions within Sri Lanka.
Note: Please fill out the request form they provide to sign the dataset agreement, after which you can obtain the dataset from the team.
Used in the KuralHub benchmark (run ID EmoTa). See the results above.
Citation
@inproceedings{thevakumar-etal-2025-emota,
title = "{E}mo{T}a: A {T}amil Emotional Speech Dataset",
author = "Thevakumar, Jubeerathan and
Thavarasa, Luxshan and
Sivatheepan, Thanikan and
Kugarajah, Sajeev and
Thayasivam, Uthayasanker",
editor = "Sarveswaran, Kengatharaiyer and
Vaidya, Ashwini and
Krishna Bal, Bal and
Shams, Sana and
Thapa, Surendrabikram",
booktitle = "Proceedings of the First Workshop on Challenges in Processing South Asian Languages (CHiPSAL 2025)",
month = jan,
year = "2025",
address = "Abu Dhabi, UAE",
publisher = "International Committee on Computational Linguistics",
url = "https://aclanthology.org/2025.chipsal-1.19/",
pages = "193--201",
abstract = "This paper introduces EmoTa, the first emotional speech dataset in Tamil, designed to reflect the linguistic diversity of Sri Lankan Tamil speakers. EmoTa comprises 936 recorded utterances from 22 native Tamil speakers (11 male, 11 female), each articulating 19 semantically neutral sentences across five primary emotions: anger, happiness, sadness, fear, and neutrality. To ensure quality, inter-annotator agreement was assessed using Fleiss' Kappa, resulting in a substantial agreement score of 0.74. Initial evaluations using machine learning models, including XGBoost and Random Forest, yielded a high F1-score of 0.91 and 0.90 for emotion classification tasks. By releasing EmoTa, we aim to encourage further exploration of Tamil language processing and the development of innovative models for Tamil Speech Emotion Recognition."
}
The study created a Tamil emotional speech database due to the lack of existing resources. It consists of 19 wave clips representing five emotions: anger, joy, fear, neutral, and sadness.
Citation
@INPROCEEDINGS{6781245,
author={Ram, C.Sunitha and Ponnusamy, R.},
booktitle={2014 International Conference on Issues and Challenges in Intelligent Computing Techniques (ICICT)},
title={An effective automatic speech emotion recognition for Tamil language using Support Vector Machine},
year={2014},
volume={},
number={},
pages={19-23},
keywords={Emotion recognition;Speech recognition;Speech;Databases;Support vector machines;Feature extraction;Mel frequency cepstral coefficient;Support Vector Machine;MFCC;EASER},
doi={10.1109/ICICICT.2014.6781245}}
Design and Development of a Multilingual Speech Corpus (TaMaR-EmoDB) for Emotion Analysis—as part of this research, they have created acted SER samples in Malayalam, Tamil, and Ravula
Citation
@inproceedings{rajan2019design,
title={Design and Development of a Multi-Lingual Speech Corpora (TaMaR-EmoDB) for Emotion Analysis.},
author={Rajan, Rajeev and Haritha, UG and Sujitha, AC and Rejisha, TM},
booktitle={INTERSPEECH},
pages={3267--3271},
year={2019}
}
Know another Tamil speech emotion dataset, or a newer version of one listed here? Suggest it or open a pull request.
This page is generated from datasets/Tamil/ in the KuralHub repository.
If you use a dataset, cite its authors. If KuralHub helped you find it, please also cite the KuralHub paper.