BEAT datasets
Opened this issue · 1 comments
zhou431496 commented
Hi, this is excellent work. I don't quite understand how to process the BEAT data. Is it using the full data set or just the English data set. Why matching can achieve semantic awareness.
YoungSeng commented
For speakers 1 and 10 of BEAT, databases were constructed separately; semantics was used with the semantic information obtained from the sentences in the 0.5s window before and after each moment after sentenceBERT, and then the cosine similarity was used to calculate the match of the text.