Why can Naive Bayes work well for high-dimensional text classification?

A. It learns nonlinear centroids for every sentence B. It ignores all feature evidence C. It requires every document to have identical words D. It estimates simple feature contributions efficiently even with many sparse word features