Hello all,
I have finetuned BERT on two different training sets for sentiment analysis. I would like to see if there was a difference caused by this. I was therefore wondering if there is a way to visualise keywords that lead to positive/negative sentiments to explore the model?
Thank you!