How To Fine Tune Bert For Text Classification Using Naive Bayes