This lecture covers the Bayesian extension of Hidden Markov Models (HMM) to automatically segment and model sequences of actions in robots trained for cooking tasks. It discusses the limitations of classical finite HMMs for segmentation, the use of Bayesian Non-Parametrics for HMMs, and the Hierarchical Dirichlet Process prior on the transition matrix. The lecture also explores the segmentation of continuous motion capture data into motion categories and the learning of complex sequential tasks from demonstration, using a pizza dough rolling case study.