This seminar will examine work on language learnability within Optimality Theory. The focus will be on the formal structure of the learning problems, the nature of algorithms that have been proposed to solve them, and the role played by the structure of linguistic theory in those proposed solutions. The empirical focus will derive primarily from linguistic data (rather than, e.g., child language production data). The course will start with an overview of relevant concepts from learnability theory. The core theory behind the learning of Optimality Theoretic constraint rankings will then be examined. Major topics in learnability to be examined include restrictive distributions ("the subset problem"), the learning of lexical representations, and computational complexity. Particular attention will be paid to the learning of phonological underlying forms, as an instance of lexical/grammatical interaction in learning.
The majority of the grade will be based upon a term paper. A good paper topic would be a detailed analysis of the learnability problems posed by some interesting linguistic phenomenon, perhaps in some grammatical system of independent interest to the student; other types of paper topics may be proposed to the instructor for consideration.
Some background in phonology and Optimality Theory will be useful for this course. The course will not presume that students have an extensive background in mathematics or computer science. NOTE: interested students from outside of linguistics are strongly encouraged to contact the instructor. If you aren't confident in your phonological background, but are interested in language learnability problems and are willing to do some extra background reading, talk to me!Course Materials