search

Vincent-ZHQ / CA-MSER

Code for Speech Emotion Recognition with Co-Attention based Multi-level Acoustic Information
130 stars 13 forks source link

About IEMOCAP Accuracy #12

Open JYeonKim opened 1 year ago

JYeonKim commented 1 year ago

Hi, I have a question. Obviously I know that the IEMOCAP dataset is multi-label, but if you look at the code, it seems that the accuracy was measured with single-label method. Maybe I misunderstood the code?

If you don't mind, can you explain it to me?

Thank you

Vincent-ZHQ commented 1 year ago

There are serval methods to use the IEMOCAP dataset. This paper is a four-class classification task and uses the final decided emotions as labels. In some papers, they use all voted emotions and treat it as a multi-label classification task. The emotion classes could also be different in different tasks. I think you have to check the original paper with the help of the raw dataset.