What is semi-supervised Naïve Bayes and how does it work?

Semi-supervised Naïve Bayes uses both labeled and unlabeled data through iterative parameter refinement.

Semi-supervised NB typically uses EM (Expectation-Maximization). First, it trains on labeled data, then assigns soft labels to unlabeled data, and iteratively refines parameters. This improves performance when labeled data is scarce but unlabeled data is abundant.