Is there any approach I can use to extract the learned lexicons with their polarities from the trained model on sentiment analysis?
I trained Arabert model to perform the sentiment analysis on Arabic reviews. Now, I want to extract the lexicon from the trained model.
If this can't be done this with deep learning models, can I train another traditional machine learning model to extract the lexicons?