How to train this model with dimensional emotion?
#9
by
Abie32
- opened
Hi, my name is abie
i just learn about machine learning, and the model that i know is for sequence classification, but this model seems like wav2vec right and i want to know, how to train this model into dimensional emotion, how to define the label beacuse the dimensional emotion should be float number, not a class
it should be regression? not a classification
Hi!
Categorical emotions can be thought as points in a unit 3D cube, you can convert the categories, e.g. relaxed to approximate points in the cube: relaxed = arousal=0.1, dominance=0.1, valence=0.99.
This model was trained with MSPpodcast that is annotated as dimensions so no conversion from categorical label was required.
