Hi, thanks for writing this very interesting paper and for providing this nice GitHub repo. :smile: I was wondering if it's possible -- and more to the point, straightforward! -- to use a SemBERT that you have trained to just get a sentence representation, which I can then use in my own neural network for classification or regression? I have a classification (or maybe ordinal regression...) task for which SRL features might be useful over and above BERT.
Hi, thanks for writing this very interesting paper and for providing this nice GitHub repo. :smile: I was wondering if it's possible -- and more to the point, straightforward! -- to use a SemBERT that you have trained to just get a sentence representation, which I can then use in my own neural network for classification or regression? I have a classification (or maybe ordinal regression...) task for which SRL features might be useful over and above BERT.