LSA.112 | Inductive Learning of Rules and Constraints

Adam Albright and Josh Tenenbaum
MW 4:50-6:30
location: 32-144

How do learners abstract over input data to arrive at a set of rules or constraints that describes what is possible in their language? This course provides an introduction to some of the techniques that are commonly used in rule induction, with an emphasis on how learners find the right level of generality to describe a category or process. The first half of the course will provide some theoretical background in rule induction, with a focus on computational modeling, and also on empirical comparison to the behavior of human learners. In the second half of the course, we will look at how inductive techniques can be integrated into models of grammar acquisition, focusing primarily on examples from phonology and morphology. A recurrent theme will be the necessity of statistical reasoning in learning linguistic structure, and the necessity of structural assumptions in statistical approaches to language.