Given the acoustic-prosodic features in emobase2010.arff, classify the emotion of each instance.
Class labels are in eval/category.txt and eval/intensity.txt. Experiment with combining the labels and trying the classification or regression task, or as multilabel learning. You can also experiment with methods for feature selection and regularization, feature normalization, speaker modeling. If you are unsure if the scope of what you want to do is appropriate, please email me a one-line "proposal" for approval.
Emotion+classification.ipynb shows an example workflow for classification.
The wav files are too large to host on github. If you are interested in doing anything with them, please email me and I will give you a semester-long account on the server where they are located.
Due date: Sunday 4/23 11:59pm.
Submission: A jupyter notebook or similar documentation with the experiments you tried (at least two), results, and conclusions.