In this section of the course, we discuss ongoing research on two kinds of “hidden structure learning” in HG: learning of underlying representations, and learning of serial derivations. Again, these models are interesting not only from the perspective of learning theory, but also for the challenges and opportunities they pose for phonological theory in general.
Papers discussed in class
Pater, Joe, Robert Staubs and Karen Jesney. In prep. Learning probabilities over underlying representations.
Staubs, Robert and Joe Pater. In prep. Learning probabilistic serial Harmonic Grammar.