Laplace smoothing adds a small count to all probability estimates to ensure no probability is ever absolute zero, which would 'wipe out' calculations in Naive Bayes.
Essential fix for zero-frequency problems.
Widely used in probabilistic modeling.