We are glad to invite you to participate in the Speech Emotion Recognition in Naturalistic Conditions Challenge at Interspeech 2025 to compare different emotion recognition systems submitted by teams all around the world. The challenge uses recordings from the MSP-Podcast corpus, which contains speech segments obtained from audio-sharing websites. The speaking turns have been perceptually annotated by at least five raters with categorical and attribute-based emotional labels.
The challenge consists of two independent tasks. Each team can participate in one or both:
Participants are encouraged to improve upon the baseline system by exploring innovative approaches, including but not limited to advanced machine learning techniques, novel feature representations, and multi-modal analyses.
Teams must register by completing and submitting the provided academic license agreement to msp-lab@utdallas.edu. Each institution must have its own signed license agreement. If you already have a license of a previous version of the MSP-Podcast corpus, you only need to request the challenge version.
Data Usage: Participants are only allowed to use the provided MSP-Podcast corpus datasets for training and development. Other publicly available pre-trained generalistic models (wav2vec, Hubert, etc) are allowed. However, the use of pre-trained models specifically trained on emotion recognition tasks is not allowed. The use of additional datasets for training is prohibited to ensure fairness.
Submissions must follow the specified file format (.csv) and be made through the designated submission platform.
Winners will be decided based on their score on the leaderboard and also on the quality of the submitted paper. The winning team(s) will receive a certificate from the challenge organizers.
Disqualification: Teams found to have violated the rules, such as using prohibited data or submitting under multiple team names, will be disqualified.
Before submission, please read and follow the instructions carefully.
Only registered team submissions will be accepts. To register please visit the
overview tab.
For information about baselines visit this link.
Each registered email is permitted a maximum of one submission per week per task. For the first submission, participants may choose any preferred team name. However, it is important to use the same team name for subsequent submissions, as any different name will result in rejection.
The submission portal will open on December 9th, 2024 and close on January 31st, 2025 (AOE).
FileName, EmoClass
MSP-PODCAST_test3_0001.wav, S
MSP-PODCAST_test3_0002.wav, D
MSP-PODCAST_test3_0003.wav, A
MSP-PODCAST_test3_0004.wav, U
...
FileName, EmoAct, EmoVal, EmoDom
MSP-PODCAST_test3_0001.wav, 5.962445394, 1.645595285, 3.277091995
MSP-PODCAST_test3_0002.wav, 5.925202743, 3.510046627, 4.902689017
MSP-PODCAST_test3_0003.wav, 5.133939009, 2.012986747, 5.79230556
MSP-PODCAST_test3_0004.wav, 2.727285967, 4.033873751, 1.566529833
...