Applying Multinomial Naive Bayes to NLP Problems: A Practical Explanation