As a computational linguist, I help computers learn to understand human language. A huge portion of human communication, thought, and culture now passes through computers. Ultimately, we want our devices to help us by understanding text and speech as a human would -- both at the small scale of intelligent user interfaces and at the large scale of the entire multilingual Internet.

My primary interests lie at the intersection of algorithms, linguistics, and machine learning. The challenge is to fashion statistical models that are nuanced enough to capture good intuitions about linguistic structure, and especially, to develop efficient algorithms to train and apply these complex models.

A central theme in my work is structured prediction -- learning to predict many interrelated variables at once. To this end, my students and I have made numerous algorithmically novel contributions to dynamic programming, belief propagation, finite-state and context-free methods, variational inference, and semi-supervised learning. We have applied these to natural language problems such as parsing, machine translation, morphology, phonology, and information extraction.

We have also been developing an innovative high-level declarative programming language, Dyna, which encapsulates many interesting efficiency tricks for such methods, and thus makes it far easier to experiment with new algorithms and models.

In general, I have broad interests and have worked on a wide range of fundamental topics in NLP, drawing on varied areas of computer science. See my research summary for more information, as well as notes on my advising style.