Jean usually records. I will check.

Recording

Thanks Jean!

Feel free to ask questions in Zoom too. I'll be monitoring them and can relay them to the speaker

Chandra has a question Brandon.

I don't see the question in chat. Chandra, could you please type that in for everyone?

Ah sure will do so!

Regarding the previous slides. My first question is how are these voice embeddings created, is this some form of conversion from speech to text and then performing word embeddings like word2vec or glove, to get the distance embeddings? My second question is regarding the loss function, if we are identifying the emotion as the class from the text, how is the model penalised? To follow up how are the annotations gathered to train this model?

Got it! Thank you

What are the baseline models here?

This is super interesting, is this paper available on arxiv? I can see it’s already published on the ACM Digital Library.

We're almost up on time, so I'll relay this to Theodora after the talk ends.

Sure, no problem 👍🏽

Quite a body of work!

Thank you, that was interesting!