The probabilistic finite-state automaton is a simple computational device
generating strings following a distribution, unlike the classical finite-state
automaton defining a language as a set of strings. This seemingly more complex
concept can be used for tasks as finding the most probable string matching a
given pattern or predicting the next character of a string prefix. The
automaton can be considered as a generalization of hidden Markov models.
The aim of to talk is to present basics on probabilistic finite-state automata
as well as some techniques of learning them from data.