The Baum Welch algorithm is a kind of EM (expectation-maximisation). It is also known as the forward-backward algorithm. Its aim is to compute the parameters (transition and emission probabilities) of an HMM (hidden Markov Model), given a learning corpus of emissions.
