The document discusses using Markov models for algorithmic composition of music. It describes Markov models as a method for probabilistically predicting the next variable in a sequence based on previous values. The document then discusses how different musical dimensions like pitch, time, and harmony can be modeled as state sequences for a Markov model to generate new musical phrases that match the style of an input corpus. It concludes by providing a link to example code for implementing such musical Markov models.