How to use individual term frequeny features with Naive Bayes classifier?

Question

I am currently working on a NLP task with Naive Bayes classifier.

My features consist out of bigrams and unigrams.

Now, to fit X_train I want to merge the top 100