Linguistics 274: Computational Psycholinguistics (Spring 2013)

This course is a reading seminar covering a variety of computational modeling approaches to human language comprehension, production, acquisition, and representation. There is a strong emphasis on probabilistic approaches: at its core, the processing of natural language involves dealing with uncertainty all the time, and in psycholinguistic research probability theory is playing a larger and larger role in modeling how people deal with this uncertainty. Over the last two decades, computational linguistics has been revolutionized by increases in computing power, large linguistic datasets, and a paradigm shift toward the view that language processing by computers is best approached through the tools of statistical inference. During roughly the same time frame, there have been similar theoretical developments in cognitive psychology towards a view of major aspects of human cognition as rational. Developments in these two fields have set the stage for renewed interest in computational approaches to human language acquisition and processing. Correspondingly, this course covers some of the most exciting developments in computational psycholinguistics over the past decade. The course focuses in particular on probabilistic knowledge and memory in language processing, covering models, algorithms, and key empirical results in the literature.

We'll meet twice a week and discuss some of the most exciting recent papers in the field. Seminar participants will take turns leading discussion of readings.

Graduate students and highly motivated, well-prepared undergraduates in linguistics, cognitive science, psychology, computer science, and any of a number of related disciplines. Postdocs and faculty are also welcome (though I won't let you out of the requirement that you lead discussion of some readings :-).

There are no formal prerequisites for this seminar, but we will be reading some fairly advanced examples of computational modeling papers, and it can't hurt you to have a good background in this area. In particular, we'll be relying on ideas from probability theory and machine learning, so some background in this area is useful. Familiarity with parsing algorithms for natural language sentences is also useful; if you've never taken a computational linguistics class, you might want to get your hands on a copy of the second edition of Jurafsky and Martin.

The requirements for participation in this seminar are that you show up, participate in discussion, lead discussion of a paper at some point during the quarter, and (if you are taking the course for credit) write a final paper (research or review) on some topic covered in the course.

There is a mailing list for this course, which you can access at https://mailman.ucsd.edu/mailman/listinfo/lign274-l. Please make sure you're signed up for it! This list is both for discussion of ideas in the class and for communications about organizational issues.

Rational analysis: Anderson 1990 Chapter 1

Speech errors: Goldrick & Daland, 2009

Emergentism: Smolensky et al., in press

My textbook-in-progress Probabilistic Models in the Study of Language – see in particular the Appendix on Directed Graphical Models