FactSet Interview Question

How does Naive Bayes work? How does smoothing help? How do you combine these ideas in a text classification system?