How to get semantic labels with face descriptions

[artpli]

Dear authors, thank you for sharing the dataset and codebase first!

I am wondering how to get semantic labels with face descriptions like ''A person is sitting and playing the piano happily'' (Figure 1 in the paper). For now, I can only find the semantic labels like ''A person is sitting and playing the piano.'' and face texts like ''Disgust'' separately. How did you combine them into a natural semantic description?

Thank you in advance for your help!

Best,
Peng